METHOD OF DETERMINING THE ORIENTATION OF A MOBILE TERMINAL-RELATED SENSOR MARK WITH SENSOR ASSEMBLY

专利摘要:
Method for determining the orientation of the path followed by a pedestrian (P), associated with a trajectory mark (RT), relative to a reference mark (RR), said pedestrian being provided with a sensor housing (BC) comprising a sensor assembly (EC) comprising at least one motion sensor, comprising the steps of: - generating data representative of the movement of the sensor housing (BC) from said sensor assembly in the reference frame (RR), and - calculating the value of a first rotation transformation operator (QRT) representative of the reference reference (RR) orientation relative to the trajectory mark (RT), so that the movement representative data thus obtained at the preceding step, in the reference reference (RR), and transformed by said first operator (QRT), have at least one characteristic of a set of characteristics representative of motion signals of a pedestrian, expressed in the pedestrian marker.
公开号:FR3015072A1
申请号:FR1362847
申请日:2013-12-18
公开日:2015-06-19
发明作者:Antoine Lemarchand；Foras Etienne De
申请人:Movea SA；
IPC主号:

专利说明:

[0001] METHOD FOR DETERMINING THE ORIENTATION OF A MOBILE TERMINAL-BASED SENSOR MARK WITH SENSOR ASSEMBLY PROVIDED BY A USER AND COMPRISING AT LEAST ONE MOTION-MOVING MOTION SENSOR The invention relates to a method for determining the orientation of the path followed by a pedestrian, associated with a trajectory mark, relative to a reference mark, the pedestrian being provided with a sensor housing comprising a sensor assembly comprising at least one motion sensor. A method for determining the orientation of the path followed by a pedestrian, associated with a trajectory marker, with respect to a reference mark, by means of a sensor assembly comprising at least one motion sensor, the pedestrian of which is provided with , can be useful in various applications such as pedestrian navigation, whether indoors or outdoors, for which it is necessary to locate the pedestrian and / or determine its trajectory. The pedestrian can practice a walking or running activity. Several techniques are known to locate a pedestrian absolutely on a plane. For example, see Robert Harle's article "A Survey of Indoor Inertial Positioning Systems for Pedestrians" published under the reference IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 3, THIRD QUARTER 2013 1281.
[0002] The most common techniques used in navigation are GNSS techniques (acronym for "Global Navigation Satellite System" in English). In optimal cases, these techniques can achieve accuracies of a few meters. The bearer of the GNNS receiver is located absolutely, at every moment. These techniques are, however, dependent on the carrier's ability to receive satellite signals. In indoor environments or "Indoor" in English, or in urban environments where several satellites may be hidden (effect called "urban canyon" in English), GNSS techniques may be inoperative or defective. In addition, they have a low power consumption balance sheet. The use of the radio signals (GSM, WIFI, etc.) transmitted and received by a mobile phone can also be exploited, including indoors to locate pedestrians absolutely in relation to the radio access points but they generate inaccurate or noisy positions (at best 100 meters for the GSM radio signal, errors of several tens of meters are possible for WIFI, including position jumps from one moment to another), they are dependent on the equipment of the place, and 10 have an adverse consumption balance. The use of a sensor assembly provided with the pedestrian, comprising at least one motion sensor, makes it possible to overcome these defects, by replacing or complementing the weaknesses of the methods conventionally used, by providing a relative trajectory, based on the instantaneous speed and heading estimates. The trajectory is then computed step by step by accumulation of elementary displacements (these techniques are known as "dead reckoning" in the English language "or" dead reckoning ") Applications can go from the big domain public in the areas of civil security or defense, in all situations including or other location systems are absent or deficient. "Dead reckoning" is very complementary to absolute location solutions.The latter provide absolute positions noisy, while the "dead reckoning" 25 provides a relative displacement.The problem of localization of people indoors has been, for several years, a very active research subject as the potential applications are numerous.One of the most promising technologies 30 ( localization by Ultra Large Band (ULB) radio technique, localization by Wifi technique, or location by Vision technique For example, the inertial measurement approach has a significant attractiveness for consumer applications because the primary measurement means (a sensor assembly comprising at least one motion sensor) are already available to users through mobile phones and tablet PCs. Other accessory devices appear on the market, also equipped with these movement measuring means, such as interactive glasses or watches worn on the wrist. The motion measurement location approach of the present application relies only on user-carried motion sensors, and is therefore independent of any infrastructure. Consumer devices such as mobile phones or other tablets, interactive glasses, watches, or other accessories already incorporate the basic sensors, namely accelerometer (A), gyrometer (G), magnetometer (M) and sensor. pressure (P), and it is then possible, without recourse to specific prior mappings and / or equipment of the place of exercise of the location (of which the other methods mentioned are dependent), to provide the information of movements during the time and thus trace the trajectory of the pedestrian.
[0003] The technique commonly used for the internal location of a pedestrian using inertial sensors (in which one often includes, in addition to the gyrometric (G) and accelerometric (A) sensors and by language abuse the magnetometer (M)) consists in applying a method called "dead reckoning" in English language or "reckoning", which estimates the current position in a reference frame (generally linked to a landmark, often confused with it, in the North sense, East, Vertical) from the previous position, to which an increment of displacement is added. For a very classic case of displacement on a two-dimensional plane (example of a ship moving on the surface of the sea, or of a vehicle moving on the surface of the earth, or of a pedestrian moving on a horizontal surface), the displacement increment is defined by a two-dimensional vector whose norm is equal to the displacement speed multiplied by the time increment since the instant of the previous position, and the angle is calculated by the heading of the moving mobile. For a boat (often referred to as "dead reckoning"), it is possible to estimate the speed of movement for example from instruments measuring the relative speed of the boat relative to water. Such an instrument may be constituted for example by a floating anchor, which is immobilized with respect to the water, and to which a rope has been hung. The floating anchor is thrown into the water by an operator, and the operator then counts the length of rope that takes place per unit of time, in order to deduce the speed of the boat relative to the water. The unwound length of rope can be identified by means of a system of nodes distributed on the rope. The operator then counts the number of nodes per unit of time (hence the speed is expressed in knots). The course can be estimated by a compass. For a land vehicle, it will be possible to estimate the speed from the number of wheel turns per second. The course can be estimated by a method similar to the case of the ship possibly complemented by the steering information given by the steering wheel. In the case of a pedestrian wearing a motion sensor system, such as those present in a smartphone or other worn accessory such as watches (wrist) or interactive glasses (worn on the head), it is common to estimate the module of the velocity vector from the measurement of the rate of operation or, which is equivalent by counting the steps taken, which can be obtained from the signals picked up by sensors sensitive to the movements that the pedestrian prints to the sensors, and a function that then transforms the rate into speed, or the number of steps in distance, according to, for example, models related to the morphology of the person. These methods are perfectible but, for example, if one has a calibration of the function performed using a test set, it is possible to obtain accuracies of a few percent.
[0004] However, in the case of the pedestrian, the course is not easy to estimate, insofar as, if the sensor system actually makes it possible to measure an orientation (and therefore a heading) of the sensor system in the reference frame ( or to fix the ideas of the terrestrial landmark), it is not possible to simply connect this orientation to the direction followed by the pedestrian. Indeed, depending on how the pedestrian carries the sensor, and there are multiple possibilities, which may also vary over time, it appears a priori that it is not possible to establish a link between the course of the trajectory and orientation of the sensor in the terrestrial reference.
[0005] A simple solution may be to ask the user to carry his sensor, ie his mobile terminal or his accessory, because the sensor is linked in motion or fast motion of the mobile terminal, constrained, for example by pointing in the direction of walking, or by fixing the sensor (ie the mobile terminal or the accessory) in a unique and known way on the body. The heading of the trajectory is then known because this heading can be deduced from the orientation of the mobile terminal. However, this method, constraining, is a brake on the adoption of location solutions in principle of inertial sensors, and secondly, there is no guarantee that the user is able to hold the deposit. Moreover, for devices that guarantee a known position on the body, such as interactive glasses equipped with movement sensors placed on the head in a known manner, or an accessory worn on the belt, it is still necessary to face the current situations in which the user turns his head in a heading that is not that of the trajectory that he follows or that the position of the accessory changes from one day to another or even during the same navigation exercise . Indeed, the position of the accessory on the belt does not alone to guarantee the relationship between the orientation of this accessory and the pedestrian heading. For watches equipped with motion sensors, the way the watch is worn remains variable from one person to another, the movements printed on the arm generate temporal variations of the watch's heading, and the port can also vary with the time for the same individual. The course of the user's trajectory can not be deduced from the course of the watch.
[0006] It therefore appears an important need to propose methods of continuous estimation of this unknown orientation of the sensor with respect to the trajectory followed by the pedestrian. It is the object of this invention to provide a method for estimating this orientation.
[0007] This problem of determining the link between sensor and trajectory, called for example "sensor-to-trajectory" or "sensor-to-traj" or "angle misalignment" in English, (angle misalignment) is very complex. complex problem is poorly studied in the literature. Most pedestrian inertial navigation systems generally assume that the orientation of the sensor relative to the path is known (for example, a sensor attached to the sternum or foot etc.). However, a number of articles and patent documents dealing with the problem have been identified below. The article "Which way am I facing: Inferring horizontal device orientation from an accelerometer signal," by Kunze K., P. Lukowicz, K. Partridge, and Begole, B., International symposium on wearable computers, 2009, assumes that horizontal accelerations (p. 149, 1. Introduction, left column) are mainly in the direction of movement. The signal is first projected in the horizontal plane using the rest periods (minimum variance on each of the axes) to calculate the vertical direction: in rest period, the accelerometric signal represents only the gravitational field, it is ie the vertical direction. This direction is then used to project the accelerometer signal in the horizontal plane.
[0008] A principal component analysis (PCA), that is to say a decomposition into eigenvectors and / or eigenvalues, is then performed on the result of the projection in order to find the direction of the displacement.
[0009] The article gives few elements, but announces five degrees of error on a course of 30m, which is important. Note in commentary of this prior art that the principle according to which the horizontal accelerations are mainly directed according to the direction of advance is not verified because, as is seen later, a pedestrian in a walking or racing situation also generates accelerations perpendicular to the direction of advancement. According to the approach of the user, there is thus a superposition of acceleration components in the direction of travel and in the direction perpendicular to walking. Another variability factor that is not taken into account by the article is the position of the sensor. Furthermore, the vertical direction identified during the rest period is no longer valid as soon as a movement that substantially changes the orientation of the sensor housing is generated. It is thus necessary to solve the problem of continuous estimation of the pedestrian heading on a terrestrial plane thanks to the sensor data, and this continuously. The article "Dead Reckoning from the pocket - An experimental study" by Ulrich Steinhoff and Bernt Schiele, Pervasive Computing and Communications (PerCom), 2010 IEEE International Conference on, 29/03/2010, focuses on a comparative and experimental study of different approaches for estimating the direction of movement, only for sensors in the pocket, using a database, comprising 8 people and 23 traces.
[0010] The principles studied to determine the direction of movement are based on two principles: - a rotational approach: the idea is that the sensor "turns" around an axis that is orthogonal to the direction of movement, approach rather focused on a sensor in the trouser pocket (the axis of rotation is that of the femur or pelvis). Note that here, the author searches for the axes of rotation of the sensor rather than directions of acceleration. It offers a totally different approach to the previous article and this highlights the lack of universal solution for the pedestrian problem, which carries a motion sensor system whose position is unknown to the body.
[0011] The article highlights the lack of robustness of the approach which is a decomposition approach in eigenvalues: PCA ("principal component analysis" in English) in 2D or 3D, filtered or unfiltered.
[0012] The idea here is that the proper accelerations are in the direction of displacement, the different methods studied simply differ in the implementation of this principle. The 3D approach looks for the 3 main components of the sensor signal, then retains the 3rd (that is, the one associated with the smallest eigenvalue) as the one that indicates the motion.
[0013] The two-dimensional or 2D approaches (PCA2D and PCA2Df) first project the signal in the horizontal plane, then perform a decomposition into eigenvectors. Here, it is the most important eigenvector (that associated with the greatest eigenvalue) that supposed to indicate the direction of walking. The article concludes that the PCA2Df approach (projection in the horizontal plane (2D) then search for the main component) with 5Hz filtering gives the best results within 5 ° of error. Here again, it should be noted that a pedestrian in a walking or running situation generates accelerations in several directions on the horizontal plane, both in the direction of travel and also perpendicularly. The combination of the two, according to the user's approach, according to the position of the sensor provides no guarantee that this resulting horizontal acceleration is always oriented in the direction of travel. US20130030754 appears to be another approach. The RRS orientation, described by equation 1, is supposed to be known thanks to the use of accelerometers and magnetometers (possibly gyrometers). The projection of the sensor landmark in the horizontal plane gives rise to a new landmark called p-frame, such that Zp_frame points upwards, and the Xp-frame and Yp-frame directions are collinear with the projection of xsensor and Ysensor, respectively. horizontal plane. In this reference, the acceleration according to Zp-frame is neglected since it can not discriminate a direction of movement in the horizontal plane, although it may be non-zero. Then it is assumed that the horizontal acceleration seen by the sensor is solely related to the displacement, and is therefore, by definition, carried by the xtraj direction. The criterion used is to look for the angle that maximizes the acceleration along that direction, which (implicitly) means that the maximum acceleration is in the direction of travel. It should be noted, however, that a pedestrian in a walking or running situation generates acceleration in the direction of the trajectory but also in directions perpendicular to the trajectory and that the combination of the two can generate results in all directions and that the result is then completely uncertain. This constitutes the heart of the principle used for determining the orientation between the sensor and the trajectory. It is also intended to correlate the accelerometer signals (or a combination of) with expected patterns corresponding to the different possible positions of the sensor (pocket, hand etc.) The search for the maximum correlation must make it possible to determine the position of the sensor (pocket, hand, etc.). However, the articulation of this technique with the previous principle does not seem very clear. No doubt the author introduces it since he finds that the previous method is not enough on its own. Moreover, it seems complex to first have to classify a sensor position on the body and then apply the method. Finally, signals from accelerometers are also used to detect footsteps, a necessary and standard technique for estimating pedestrian movement. WO2012141811 discloses an estimate of the position based on a selection of possible positions, for each of which a likelihood is calculated. The likelihood is calculated from the data of the accelerometers, light sensors, and more generally all the available sensors, for example: the on / off situation can be detected by using the variance of the standard of the accelerometers; the pitch and / or roll angles, calculated from the averaged accelerometer data, give indications on the attitude of the device; the angle θ = direction of motion, can be calculated by decomposition in eigenvalues (similar to the method of WO2012158751A1, without all the refinements to suppress the transverse accelerations); - the use of MFCCs derived from speech recognition can help classification; and the use of the optical sensor makes it possible to determine the number of occluded faces.
[0014] This multi-sensor fusion is complex, it requires a classification stage itself subject to error to produce results. The document WO2012158751 deals essentially with a decomposition into eigenvectors and / or eigenvalues of the accelerometer signal, except that the transverse accelerations are suppressed by a pretreatment comprising the following steps consisting in: separating the accelerations in horizontal and vertical accelerations; - suppress transverse accelerations by summing the accelerations of the "left" and "right" steps to obtain a zero resultant; 15 - if the sensor is not centered, that the resultant is non-zero, correlate the vertical accelerations with the horizontal delayed / advanced because it has been noticed that these are out of phase of +/- 72. The result is strong for so-called "displacement" acceleration and low for transverse accelerations, which are, however, also accelerations related to the movement, that is to say the movement of the pedestrian in a walking or running situation; and - the angle misalignment (MA) is finally calculated by decomposition into eigenvalues. This method is based on the principle that the main acceleration measured is in the direction of travel and proposes a mode of selection of accelerations according to the direction of the march which is subject to error. WO201316875 provides a description of a complete pedestrian navigation system, centered on the detection of the direction of movement. It also provides the use of radio (WIFI, GPS), maps, particle filter for data fusion. The determination of the direction of movement is close to US20130030754 and consists of finding the main direction of acceleration. The idea is to extract the high-frequency components of the signal that are deemed to contain the eigen-accelerations while the low-frequency components contain the gravitational field. This operation is carried out by a simple high-pass filtering of the accelerometer signals. At this point, the signals contain the accelerations in the 3D space, that is to say of dimension 3 × N, where N is the number of measurements collected. It is then said that the only interesting components are horizontal accelerations, that is to say those that are orthogonal to the z axis of the reference frame, however it is not said how these components are extracted (from many techniques are possible to project the signals in the horizontal plane). Moreover, by definition of the low frequency components, the method obtains an average direction of the vertical, which is not applicable to the whole movement. The extraction of horizontal components does not seem possible, as soon as the movement includes substantial rotations. The extraction of horizontal components leads to a matrix 2 x N.
[0015] From this matrix, we calculate the energy of the horizontal components of the accelerometer signal, then this energy signal is filtered. It is then assumed that the energy is maximum in the direction of movement. This hypothesis is subject to caution, because as we have introduced previously, and as will be exposed again with more details, a pedestrian in a walking or racing situation generates both accelerations in the direction of the trajectory, but also accelerations in the direction perpendicular to the trajectory. Here again, the hypothesis on which the process is based is fragile.
[0016] We then look for the angle rotation that maximizes the energy in the first direction, (which amounts to diagonalising the energy matrix, that is, performing a decomposition into eigenvectors) by solving a criterion.
[0017] This is an analytical solution for calculating the angle.
[0018] The main weakness of all these approaches lies in their robustness, especially in the hypothesis that the measured accelerations are borne by the direction of movement, whereas the accelerations generated by the movement of a pedestrian in a walking or running situation generate both transverse and longitudinal accelerations. Furthermore, the robustness of the treatment at the position of the sensor is not covered, we find that depending on the position of the sensor, particularly on limbs such as arms or legs, the principles of calculating the direction of the trajectory are not the same. Indeed, whatever the method used or the name given to it, all these approaches are based on the fact that the measured acceleration component is collinear with the direction of movement. This hypothesis seems likely for the foot movements of the pedestrian. However, it is possible to wonder about how this acceleration spreads from the feet to the other possible positions of the sensor: hands, pockets, chest ... it appears indeed that these points of the body have directions of movement. accelerations much less marked than those of the feet. We will see in the rest of the present application that the movement of the body is characterized by accelerations in the direction of walking but also in the direction perpendicular to walking. The power of these accelerations depends in particular on the approach of the user and also on the position of the motion sensor system on the body. In addition, other positions of the body are affected by movements unrelated to the direction of movement and can cause relatively large accelerations, this is for example the case of pendulum arm movements. Finally, again for reasons of robustness, the implementation of these principles involves processing a sufficiently long signal duration in order to attenuate the effect of the punctual parasitic phenomena and to extract an "average" effect. One problem is that the notion of duration is contradictory with that of real time, which is central to the applications envisaged. Vertical directions are sometimes found by low-pass filtering (in order to extract the value of the direction of the gravity field), and as soon as the movements printed on the sensor system include substantial rotational movements, this mean direction can no longer to be used. This is true for cyclical walking movements. On the other hand, for example, if a user changes the position sensor, the new position can only be accurately determined after a duration T such that the accelerometer signal collected during this period contains several step cycles. The duration of a step being of the order of one second, one imagines that durations of ten seconds are easily reached (analysis of the clean accelerations on ten steps) to possibly see more, according to the compromise retained between the latency of the system and the desired accuracy. The present invention addresses this problem by an innovative approach where the "sensor-to-traj" orientation is continuously estimated, by identifying the unknown orientation between the sensor (or set of sensors) 20 and the pedestrian path. Also, it is proposed, a method for determining the orientation of the trajectory followed by a pedestrian, associated with a trajectory mark, with respect to a reference mark, said pedestrian being provided with a sensor housing comprising a sensor assembly. comprising at least one motion sensor, comprising the steps of: generating data representative of the movement of the sensor housing from said sensor assembly in the reference frame, calculating the value of a first rotational transformation operator representative of the orientation of the reference mark with respect to the trajectory mark, such that the representative data of the movement thus obtained in the preceding step, in the reference mark, and transformed by said first operator, have at least one characteristic of a set of characteristics representative of pedestrian movement signals, expressed in the r pedestrian.
[0019] In one embodiment, the reference mark is a terrestrial mark (linked to the earth, on the ground), and said generation of the data representative of the motion of the sensor case is obtained from said set of sensors in the reference mark by application of a second rotation transformation operator for determining the orientation of the trajectory mark in said terrestrial reference. Thus, the invention makes it possible, when the reference reference (RR) is linked to the Earth, to always know the orientation of the reference frame linked to the Earth relative to the trajectory mark (RT) of the trajectory followed by a pedestrian equipped with a sensor box, and to be able to determine the pedestrian's heading in the reference linked to the Earth, in order to apply the techniques of "dead reckoning" in English. In another embodiment of the invention, when the reference mark is linked to the sensor housing, the invention makes it possible to always know the orientation of a sensor housing with respect to the pedestrian (trajectory mark) equipped with a terminal including the sensor housing, and to be able to activate the terminal differently, depending on this orientation. The terminal may be a mobile terminal type mobile phone or tablet, game station, interactive glasses, watch or other accessory with motion sensors and worn by the pedestrian. Also, data from the sensor assembly represented in the (terrestrial) reference frame are available thanks to the second rotation transformation operator, and this at any moment, and the invention then makes it possible to determine the orientation of the reference mark. terrestrial reference in relation to the mark of the trajectory followed by the pedestrian. This makes it possible to feed dead reckoning methods and thus provide the course of the trajectory actually followed by the pedestrian in the terrestrial reference. The sensor housing carried by the pedestrian can be oriented in any way with respect to the pedestrian. Thus, there is a method capable of providing the pedestrian heading in the terrestrial reference, robust at different positions of the sensor on the pedestrian, and able to adapt to changes in position.
[0020] According to one embodiment, the reference mark and the trajectory marker further comprise a common axis, so that the first operator of transformation in rotation is reduced to a transformation operator in rotation along the common axis.
[0021] With this common axis, it is simpler to determine the first operator, since it is then no longer necessary to estimate a single unknown angle between the two reference marks and trajectory around this common axis. The first operator is reduced to a rotation along a single axis. The methods for calculating the pedestrian's heading are simplified. In one embodiment, the common axis is further oriented in the direction of Earth's gravity, so that the first rotational transformation operator is reduced to an operator of rotational transformation along a direction axis of rotation. Terrestrial gravity. Also, several advantages can be described. In the first place, reference is made to references conventionally used by those skilled in the art, that is to say marks for which the vertical direction according to terrestrial gravity is used as one of the axes of the reference marks. exploited in the invention. Secondly, these markers comprising a vertical axis are directly in conformity with the markers in which the characteristics of the movements of a pedestrian in a walking or running situation are conventionally described and it will then be easier to identify the conformity of the representative data of the movement of the pedestrian with respect to these characteristics, which constitutes the heart of the invention. In the third point, a single angle must be determined in order to completely determine the value of the first operator which makes the method simpler whereas in the general case three angles or three parameters depending on the mode of representation of the selected rotation operators must be determined. In fourth point, this angle will be directly that of the trajectory in the terrestrial frame, since the trajectories and thus the caps sought in the great majority of the cases of exploitation of the invention, must be represented on maps or plans representing the geographical data in a horizontal plane, the vertical axis being perpendicular to these maps or to these plans. According to one embodiment, the orientation of the sensor housing in the trajectory marker (or pedestrian marker) is determined by the composition of the second operator with the first operator.
[0022] Also, it is then possible to determine the orientation of the housing in the pedestrian mark and activate functions automatically according to this orientation. It can then be determined according to this orientation if functions of a mobile terminal, interactive glasses must be activated or not. For example if it is determined that the user wearing interactive glasses looks in the direction of walking, it may be relevant to return some information to him while when looking in a different direction, it should be provided to him other information. Similarly, if the screen of a terminal such as a mobile phone is oriented with a certain orientation with respect to the pedestrian, certain display functions can be activated. In one embodiment, a so-called "central attitude" function or an "Inertial Measurement Unit" language abuse function is performed which provides the value of the second operator which makes it possible to transform the data directly derived from the sensors and thus identified in the reference of the sensor housing, in the reference reference linked to the Earth.
[0023] Also, the second operator is calculated directly from the data from the motion sensors present in the sensor housing and it is not necessary to use other sensors to estimate the value of the second operator. The device is thus completely autonomous.
[0024] According to one embodiment, the central attitude function calculates the second operator from a combination of data provided by accelerometric and / or gyrometric and / or magnetic inertial motion sensors present in the sensor housing. Also, we do not depend on any other external equipment and the same sensors of the sensor case can be used to determine the second operator and this at any time which allows to determine the movement data in a reference frame related to Earth.
[0025] In one embodiment, when the pedestrian practices a walking or running activity, said set of characteristics representative of movement signals of the bust, thorax or pedestrian pelvis represented in a pedestrian marker defined by the axes of the anteroposterior pedestrian, mediolateral, and vertical, said pedestrian mark then being linked to the trajectory mark comprises the following characteristics: the signal due to the movement in translation along the medio-lateral axis has essentially power at the stride rate; the signal due to the translational movement along the anteroposterior axis has essentially power at the pace of the step; the signal due to the translational movement along the vertical axis has essentially power at the pace of the step; the signals due to the movements in translation along the vertical axis and along the antero-posterior axis have a substantially constant phase shift; the step rate is substantially double the stride rate; the signal due to the rotational movement along the medio-lateral axis has essentially power at the pace of the step; the signal due to the rotational movement along the anteroposterior axis has essentially power at the rate of the stride; and the signal due to the rotational movement along the vertical axis essentially has power at the rate of the stride.
[0026] Also, for positions of the sensor housing for which the movements measured by the motion sensors are related to those of the chest, chest or pelvis of the pedestrian, it is possible to design methods or methods combining one or more characteristics.
[0027] The exploitation of a characteristic among the set of characteristics makes it possible to design simple methods for estimating the transformation operator in unknown rotation, which is the subject of the invention, whereas the combination of several characteristics makes it possible to design methods. more complex and robust because combining several criteria. In addition, it is possible to select one or more criteria depending on the nature of the motion sensors available in the sensor housing. The methods can notably exploit the anteroposterior accelerations, the medio-lateral accelerations, which are adapted if the sensor box has an accelerometer, the rotations along the medio-lateral axis, the rotations along the anteroposterior axis, which are adapted if the Sensor housing has rotation sensors such as gyrometers. Among the characteristics, it should also be noted that these movement characteristics occur at particular frequencies (those of stride or step). It is then possible to finely select the characteristic (s) to be used in order to determine the operator representative of the orientation of the reference mark relative to the pedestrian mark, which is itself equal to the trajectory marker. According to one embodiment, when the pedestrian practices a walking or running activity, said set of characteristics representative of movement signals of a free limb of the pedestrian (such as arms or legs) represented in a pedestrian mark defined by the axes of the anteroposterior pedestrian, medio-lateral, and vertical, said pedestrian mark being linked to the trajectory mark comprises the following characteristics: the signal due to the translational movement along the Antero-Posterior axis essentially presents power at the rate of stride; the signal due to the translational movement along the vertical axis has essentially power at the pace of the step; the signal due to the rotational movement along the medio-lateral axis has essentially power at the rate of the stride; the rotation signal due to the rotational movement along the vertical axis essentially has power at the rate of the stride. Thus, for positions of the sensor assembly for which the sensed movements are related to those of a free limb of the pedestrian, including his arms or legs, it is possible to devise methods of estimating the operator of the rotational transformation representative of the orientation of the reference mark relative to the pedestrian mark, that is to say, the trajectory mark, by combining one or more of the listed characteristics. The combination of several criteria makes the trajectory estimation method more robust. These methods can notably exploit the anteroposterior accelerations, which are particularly suitable when the acceleration measurement is available, the rotations along the medio-lateral axis, which are particularly suitable when a rotation sensor, such as a gyrometer, is available. also knowing in addition to which frequencies these movements are located and / or simultaneous combinations of all these properties. In one embodiment, said one or more characteristics are selected from the set of characteristics of the movements of the pedestrian bust as described previously or in the set of characteristics of the movements of a free pedestrian limb as described previously, to from an indicator characterizing the nature of the mechanical connection between the sensor housing and the pedestrian.
[0028] Thus, it is possible to implement the right set of characteristics corresponding to the indicator and to find the path of the pedestrian. When the sensor housing is only integral with the bust, such as interactive glasses, we will use exclusively characteristics related to the movements of the bust of the pedestrian. When the sensor housing is only integral with the pendulum arm, as for a watch equipped with motion sensors, only characteristics related to the pedestrian arm will be used. In the case where the two situations are possible, one or the other set of characteristics will be used, according to an indicator representative of the port of the sensor case. According to one embodiment, said data representative of the movement of the sensor housing in the reference frame are generated from at least one accelerometer with at least two measurement axes, and said characteristic is that the acceleration signal due to the walking / running along the main direction of the trajectory or along the anteroposterior axis essentially presents a power peak at the pace of the step. In other words, a rotation transformation operator according to the vertical QRT axis is calculated so that the acceleration thus given in the horizontal plane of the reference mark and transformed by said rotation operator along the vertical axis , essentially presents a power peak at the rate of the pitch along the main axis of the trajectory.
[0029] Thus, only one remarkable characteristic of the walking signal of the pedestrian is exploited, the analyzed signal is restricted to the frequency of the step and makes the method more precise, since it is known that the acceleration signal of a pedestrian, at the frequency of the no is essentially in the direction of the trajectory, and the method remains simple to implement.
[0030] In a variant, said data representative of the movement of the sensor housing in the reference reference mark are generated from at least one accelerometer with at least two measurement axes, and said characteristic is that the acceleration signal due to walking / running in the direction perpendicular and horizontal to the main direction of the trajectory or in other words along the medio-lateral axis essentially presents a peak of power at the rate of the stride. In other words, the transformation operator in rotation along the vertical axis QRT is calculated so that the acceleration thus given in the horizontal plane of the reference mark and transformed by said rotation operator along the vertical axis essentially a peak of power at the pace of the stride along the mid-lateral axis of the pedestrian (that is to say perpendicular and horizontal to the main axis of the trajectory).
[0031] Thus, only one remarkable characteristic of the walking signal of the pedestrian is exploited, the signal analyzed is restricted to the frequency of the stride and makes the method more precise, since it is known that the acceleration signal of a pedestrian, at the frequency The stride is essentially in the mediolateral direction and the method remains simple to implement. In one embodiment, the rotational transformation operator along the vertical axis is determined so that the phase shift between the acceleration due to walking / running measured along the vertical axis and the acceleration due to walking / stroke and transformed by said operator along the axis AP is between 0 and Tr, and is particularly 7/2.
[0032] Thus, it is possible to determine the direction of the walk and not just the direction of the walk. According to one embodiment, the operator for transformation in rotation along the vertical axis is determined from the amplitudes, at the pitch frequency, of the two horizontal components of the acceleration signal in the reference frame. Also, the angle of rotation transformation operator is determined from the two horizontal components of the acceleration signal 20 in the reference frame, which allows on the one hand to find an angle that maximizes the amplitude of the the acceleration at the pitch frequency in the direction of the walk and minimizes the amplitude of the acceleration at the pitch frequency in the direction perpendicular to the step. It is also possible to provide a confidence factor, for example by determining a ratio between the magnitude of acceleration thus found in the direction of travel and the magnitude of acceleration thus found in the direction perpendicular to walking. When this ratio is large compared to 1, the confidence factor in the estimated angle is large, when it is close to 1 (and a fortiori smaller than 1) the confidence factor is small. Other confidence factors can be formed, depending on the amplitudes found in the AP and ML directions. For the analysis of the signals at the step frequency, the power or the amplitude of the signals transformed by the transformation operator in rotation of the reference mark to the pedestrian marker will be entirely found along the axis AP (and thus little amplitude or power along the ML axis), the better will be the confidence.
[0033] In one embodiment, the rotational transformation operator along the vertical axis is determined from the amplitudes, at the stride frequency, of the two horizontal components of the acceleration signal in the reference frame. Thus, the angle of rotation transformation operator is determined from the two horizontal components of the acceleration signal in the reference frame, which allows on the one hand to find an angle that maximizes the amplitude. from the acceleration to the frequency of the stride in the direction perpendicular to the walk and minimizes the amplitude of the acceleration at the frequency of the stride in the direction of the walk. It is also possible to provide a confidence factor, for example by determining a ratio between the amplitude of acceleration thus found in the perpendicular direction of the step and the amplitude of acceleration thus found in the direction of travel. When this ratio is large compared to 1, the confidence factor in the estimated angle is large, when it is close to 1 (and a fortiori smaller than 1) the confidence factor is small. Other confidence factors can be formed, depending on the magnitudes found in the AP and ML directions. In the case where the signals are analyzed at the frequency of the stride, the power or the amplitude of the signals transformed by the transformation operator in rotation of the reference mark to the pedestrian marker is entirely found along the axis ML (And therefore can amplitude or power along the axis AP), the better will be the confidence. According to one embodiment, said data representative of the motion of the sensor housing in the reference frame are generated from at least one accelerometer to at least two measurement axes, and the two preceding characteristics are combined: (i) the acceleration signal due to walking / running along the main direction of the path (or along the antero-posterior axis) essentially presents a power peak at the pace of the step, and (ii) the acceleration signal due to the walking / running in the direction perpendicular and horizontal to the main direction of the trajectory or in other words along the medio-lateral axis essentially presents a peak of power at the rate of the stride. It is naturally possible and advantageous to simultaneously exploit the two amplitudes of the accelerations in the horizontal plane at the frequency of the stride and the frequency of the march, by combining the criteria previously established for the one and the other characteristic of the movements at the frequency of the stride and the frequency of walking. The optimal angle is found when the 2 horizontal accelerations at the pitch frequency (AccX (Step) and AccY (Step)) and the 2 horizontal accelerations at the frequency of the stride (AccX (stride) and AccY (stride)), transformed by the transformation operator in rotation along the vertical axis are transformed respectively for (AccX (Pas) and AccY (Pas)) into 2 horizontal components at the pitch frequency (AccAP (Pas) and AccML (Pas)) such that the first presents a maximum amplitude according to the Anteroposterior axis and the second a minimum amplitude according to the axis Medio Lateral and for (AccX (stride) and AccY (stride)), in 2 horizontal components with the frequency of the stride (AccAP (Stride), AccML (Stride)) such that the first has a minimum amplitude along the Anterior-Posterior axis and the second a maximum amplitude along the Medio-Lateral axis. It is also possible to form confidence criteria by calculating, for example, ratios between the amplitudes AccAP (not) and AccML (not) which must be large in front of 1 for a high confidence in the estimated angle, and the amplitudes AccML (Stride ) and AccAP (Stride) which must be large in front of 1 for a high confidence in the estimated angle. Other confidence criteria can be built, which measure respect for the outstanding characteristics that are exploited. The more the characteristic is respected, the better the confidence in the estimation of the operator in rotation. In the rest of the description, other criteria are presented. In other words, the rotation transformation operator according to the QRT vertical axis is calculated so that the two characteristics are exploited, namely: (i) the acceleration thus given in the horizontal plane of the reference mark; reference and transformed by said rotation operator along the vertical axis essentially has a power peak at the rate of the step along the anteroposterior axis of the pedestrian (that is to say along the main axis of the trajectory) and (ii) the acceleration thus given in the horizontal plane of the reference mark and transformed by the said rotation operator along the vertical axis essentially presents a power peak at the cadence of the stride along the mid-lateral axis of the pedestrian ( that is perpendicular and horizontal to the main axis of the trajectory) According to one embodiment, the frequency of the steps or the stride is determined from the acceleration signal along the vertical axis. cal. Thus, irrespective of the position and orientation of the sensor housing and thus of the sensor housing port by the pedestrian, it is possible to determine the pitch frequency, since the acceleration signal according to the vertical is essentially characterized by power at the frequency of steps. In one embodiment, the amplitude of the pitch frequency acceleration signal in the horizontal plane of the reference mark is determined by filtering the acceleration signal in the horizontal plane of the reference mark, said filter being characterized by its impulse response taken equal to the acceleration signal in the vertical direction in the reference frame. This technique is called matched filtering, intercorrelation, or synchronous detection. Also, this embodiment makes it possible to calculate perfectly the amplitude of the acceleration signal in the horizontal plane at the pitch frequency and provides the phase shift between the acceleration signal according to the vertical (which has essentially power at the frequency of steps) and according to the horizontal at the frequency of steps. In one embodiment, said data representative of the movement of the sensor housing in the reference frame are generated from at least one gyrometer with at least two measurement axes, and for which at least one of said one or more characteristics 35 is that the rotational speed signal due to walking / running along the mid-lateral axis has essentially a peak power at the rate of the stride. In one embodiment, when the movements printed to said sensor assembly are essentially due to the movement of the pedestrian thorax-pelvis assembly, at least the characteristic that the translational motion signal at the not essentially presents power along the antero-posterior axis, and when the movements printed to said sensor assembly are essentially due to movement of a free pedestrian limb, at least the characteristic that the motion signal rotation at the frequency of the stride presents essentially power according to medio lateral. The invention will be better understood from the study of some embodiments described by way of non-limiting examples and illustrated by the appended drawings in which: FIG. 1 schematically illustrates the pedestrian system, with the various reference marks, for to better understand the process according to one aspect of the invention; FIG. 2 schematically illustrates the various references used; - Figure 3 schematically illustrates a pedestrian and its associated pedestrian mark; Figure 4 schematically illustrates the different positions of the sensor housing; - Figure 5 schematically illustrates a period of walking or stride, a stride being composed of two steps; Figure 6 schematically illustrates an embodiment of the invention; FIG. 7 schematically illustrates an unknown angle of rotation between the 2D mark formed by the XR and YR axes and the mark formed by the ML and AP axes; FIG. 8 represents the time signals of the unknown angle, according to step 2, according to one aspect of the invention; FIG. 9 represents the time signals of the unknown angle, according to step 3, according to one aspect of the invention; FIG. 10 illustrates a path of rectangular shape, according to one aspect of the invention; FIG. 11 illustrates an example of time signals for a route of FIG. 10; FIG. 12 illustrates some periods of the acceleration signal, estimated in the pedestrian marker, according to one aspect of the invention; FIG. 13 illustrates a result according to the same formalism as that of FIG. 11, with an embodiment of the path in which the sensor is held in "landscape" mode; and FIG. 14 represents, according to the same formalism as that of FIG. 12, a few periods of the acceleration signal in the pedestrian marker.
[0034] In the set of figures, the elements having the same references are similar. FIG. 1 schematically illustrates a pedestrian system and its operation according to one aspect of the invention, and in particular the different reference marks used and their relations. A sensor housing BC comprises an EC sensor assembly provided with at least one motion sensor. A housing mark RB is associated with the sensor housing BC. A reference mark RR is also determined. A rotation transformation operator between the reference box RB and the reference reference RR is noted QBR. This operator is likely to evolve in the meantime. A pedestrian P is provided with a pedestrian marker R. A trajectory marker RT is associated with the trajectory followed by the pedestrian P. A QPT operator is defined between the trajectory marker and the pedestrian marker. This operator is defined by convention of the pedestrian landmarks Rp and trajectory RT.
[0035] The pedestrian P and the sensor housing BC are of course linked by a mechanical coupling.
[0036] The present invention makes it possible to determine the orientation of the trajectory followed by the pedestrian P, associated with the trajectory marker RT, with respect to the reference reference RR. This orientation is noted QRT and is likely to evolve over time.
[0037] Depending on the choice of the reference reference RR, the applications of the invention may be different. When the reference reference RR is linked to the sensor housing BC, the invention makes it possible to always know the orientation of the sensor housing BC with respect to the pedestrian P (path marker RT) equipped with a terminal comprising the sensor housing BC, and to be able to activate the terminal differently, depending on this orientation. The terminal can be a mobile terminal type mobile phone or tablet, game station, interactive glasses, or wrist strap, ...
[0038] The invention also makes it possible, when the reference reference RR is linked to the Earth, to know the orientation of the reference linked to the Earth in the RT trajectory marker and thus to be able to determine the pedestrian heading, an essential component in order to be able to apply the dead reckoning techniques in the English language.
[0039] The method comprises the steps of: - generating data representative of the movement of the sensor housing BC from said sensor assembly in the reference reference RR, - calculating the value of a first rotation operator QRT representative of the orientation reference reference RR relative to the trajectory mark RT, such that the data representative of the movement thus obtained in the preceding step, in reference reference RR, and transformed by said first operator QRT, have at least one characteristic of a set of characteristics representative of pedestrian movement signals, expressed in the pedestrian marker.
[0040] The general objective is to determine the direction of the trajectory followed by a pedestrian, at each moment, said trajectory being characterized by a trajectory mark at the current instant noted RT (see FIG. 1), in a reference frame (reference RR, often the landmark). The desired direction is provided by the orientation of the trajectory marker in the reference frame for the current moment, that is to say by the data of a QRT rotation transformation operator between the two marks. The data of successive QRT operators in time, associated with the data of successive velocities in the time of the pedestrian, allows for example to trace the path of the pedestrian. This technique of trajectory estimation by the data of a heading and a speed is commonly called "Dead Reckoning" or "Deduced Reckoning" in the English language, often noted DR. We are interested here in the problem of estimating the pedestrian's heading at every moment, without being interested in speed. The complexity of the problem arises from the fact that we do not have a direct measurement of the pedestrian's heading in the reference reference RR, but more indirectly measurements from motion sensors of a housing BC carried by the pedestrian. Indeed, the position and orientation of the BC housing on the pedestrian are not known. Thus, even if one can consider that the orientation of the housing BC relative to the reference reference RR is known, or more probably or practically, can be obtained from the motion sensors present in the housing, it is not it is not possible to determine the orientation of the pedestrian (and therefore of his trajectory) in the reference reference RR. It is considered that the pedestrian is provided with a sensor housing BC comprising motion sensors and that it moves while walking or running with the housing. The case BC comprises, for example, accelerometer type sensors and / or magnetometers and / or gyrometers, which are currently available with three measurement axes. Thus an accelerometer A directly provides the acceleration field which can be represented by a vector with three components. The same goes for a magnetometer M, which directly supplies the magnetic field, which can be represented by a three-component vector. Similarly, a gyrometer G provides rotational speeds along its three axes, the rotational speed can also be represented by a three-dimensional vector. These three sensors are commonly available and are now standard on mobile phones or smartphones, or touch pads. The sensors provide their measurements in the reference of the sensor or reference of the BC box. The most iconic wearable device is a mobile phone or tablet or even a laptop. It can also be any other object equipped with motion sensors, such as interactive glasses, or accessories worn by the user. Mobile phones, tablets or computers now commonly carry such sensors. The interactive glasses can be easily equipped with such sensors. Electronic accessories worn by a person are also commonly equipped with motion sensors, especially those intended for monitoring the physical activity of people. Specialized radio terminals for isolated workers are also part of the devices capable of carrying such sensors. The invention applies to any electronic device carried by a person and equipped with at least one motion sensor. All of the electronic devices mentioned are also equipped with computing means, connection means and data communication to computer networks. The calculation implementing the invention can be entirely performed on board the electronic device, without the need for communication with a network.
[0041] It is one of the advantages of the invention to be able to estimate each moment, only from the measurements of motion sensors embedded in the device, the heading of the person who wears it to estimate its trajectory. We are thus independent of any infrastructure.
[0042] To set the ideas we can imagine an emblematic scenario of the present invention, wherein the pedestrian is equipped with a mobile phone or smartphone comprising or being considered a BC sensor housing, equipped with motion sensors, and the lens consists of estimating the path of the pedestrian using only motion sensors present on board the BC sensor or smartphone. The data, at each moment, of the pedestrian velocity vector in the reference frame makes it possible to apply the technique of Dead Reckoning. The velocity vector can be decomposed into a module (the speed of movement of the pedestrian in the reference frame) and the data of its orientation in reference reference RR. The main focus is on the determination of this orientation information which is very complex, as the pedestrian is free to carry his mobile phone in different ways and so the orientation of the mobile phone does not determine the course. of the trajectory. In the majority of industrially interesting cases, the pedestrian moves on horizontal planes with respect to a terrestrial landmark. In this type of scenario, providing the orientation of the pedestrian velocity vector in the RR reference mark is then limited to providing a single heading angle of its trajectory on the horizontal plane of movement.
[0043] The mainly considered motion sensors (A, G and M) provide three-dimensional measurement vectors. The sensors are rigidly connected to the BC sensor housing. It can therefore be considered that the sensors thus deliver their vector measurements in a same frame, linked to this sensor housing BC. The person skilled in the art knows the methods which make it possible to correct, if necessary, the possible misalignments between the axes of the sensors themselves and thus which provide the measurements of the sensors in this same and unique reference box RB. For small misalignments, the invention can still apply.
[0044] As mentioned above, techniques are known to provide the orientation of the case BC in the reference frame RR, when this reference frame is defined as linked to the Earth. It must be noted, for example, that the sensors A, G, M, conventionally present in smartphones make it possible to constitute an attitude center or "Inertial Measurement Unit" of acronym IMU in English language, which, for example, by sensor data fusion technique, provides the orientation of the reference mark. RB case relative to a reference reference linked to the Earth, that is to say the QBR rotation transformation operator. Many combinations are possible between the sensors A, G, M in order to arrive at a satisfactory estimate of QBR. If the housing is equipped with very powerful sensors, for example a low drift gyrometer and very good calibration, it is possible to estimate at each instant this orientation with respect to a landmark only on the basis of the gyrometer on a horizon of several tens of minutes, even hours. Note, however, that the unknown issue of the pedestrian's heading remains, because the orientation of the sensor housing, even precise, does not determine the course of the trajectory. A rotation transformation operator can be represented in different forms, a rotation matrix, a quaternion, several rotation matrices operated in series, for example according to the Euler or Cardan conventions. In Figures 1 and 2, a view of the different marks and transformation operators between the marks. The data of an operator transforming into rotation between two marks makes it possible to transform a vector from one marker to another, in one direction and in the opposite direction. Those skilled in the art can use the appropriate formalism to represent these transformation operators in rotation. A rotation transformation operator is completely defined by the data of the axis of rotation (given by a unit vector with two independent parameters) and the angle of rotation is three independent parameters. To complete the conventions of marks and notations, it is considered that the pedestrian is provided with a pedestrian mark Rp according to Figure 3. Any other reference is naturally possible, and does not restrict the generality of the present description. However, it is simple and we will advantageously use such a formalism to exploit a particular pedestrian marker, but commonly used. The pedestrian is provided with a pedestrian marker denoted Rp defined by a first anterior-posterior axis AP, a second medial-lateral axis ML and a third vertical axis VT (as illustrated in FIG. 3). The triaxle (ML, AP, VT) forms the pedestrian mark Rp. During a "normal" walking or running activity, it may be considered that the anteroposterior axis AP is directed in the direction of the trajectory (cf. 2). Thus, for example, knowing the orientation of the trajectory marker RT in the reference reference RR is equivalent to knowing the orientation of the pedestrian marker Rp in the reference reference RR, since naturally the pedestrian has a directed trajectory according to the anteroposterior axis AP. The two marks: pedestrian marker RP and path mark RT are therefore equivalent, possibly to a constant rotation operation close. We note this transformation operator in rotation QPT. The transformation operator in rotation QPT is constant and known. For example, according to the notations and conventions of Figure 2, it is equal to the identity: (AP, ML, VT) = (XT, YT, ZT). We can refer to Figure 1 which shows, in block diagram form, the various references and the links between the various references useful for understanding the present invention.
[0045] To solve the problem of determining the orientation of the path of the pedestrian P, if the mechanical coupling between the sensor housing BC and the pedestrian P was known (by the data of a rotation operator between the housing mark RB and the pedestrian marker RP that we note QPB), we could easily find the QRT orientation of the trajectory by composition of the orientation of the QBR box BC in the terrestrial reference RR with this rotation operator QPB. However, it is assumed that this orientation is unknown, because the pedestrian can wear his case (for example his smartphone) in various ways unknown a priori, and it can also change the mode of porting in use. The rotation operator QRT is calculated by the composition of the rotation operator QPB with the rotation operator QBR. As explained in the previous paragraph and further recalled later, methods well known to those skilled in the art allow estimate the QBR orientation of the sensor housing BC in the reference reference RR. The problem of estimating the QRT orientation of the trajectory in the RR reference mark is therefore equivalent to that of determining the orientation of the casing BC with respect to the trajectory (or pedestrian mark RP, which are equivalent). , QPB. If we know how to determine the rotation operator QRT then we can calculate the rotation operator QPB, and vice versa. The present invention solves both problems. The determination of the path of the pedestrian P has all kinds of utilities to locate it at any time, by application of dead reckoning techniques.
[0046] The invention is therefore of major interest, especially for cases in which the conventional principles of location are not operational. The most common conventional system is the GPS ("Global Positioning System" in English) based on the exploitation of the so-called GNSS principle or GLONASS (for "global navigation satellite system" in English). GPS makes it possible to provide an absolute location every moment from the reception of signals generated by satellites in orbit around the Earth. It allows to locate vehicles, pedestrians, ....
[0047] However for situations where the GPS signal is not satisfactory, or absent, the satellite location is not correct or unavailable. Thus situations in the city for which the presence of buildings gene the reception GPS are qualified effect canyon or "Canyon Effect" in English language. The mobile to be located is no longer in direct view of a sufficient number of satellites and its location is then no longer satisfactory. Worse, in indoor or indoor situations in English, GPS signals are not available and no location is possible.
[0048] The invention allows to feed or provide input data an estimator of the path followed by the pedestrian P by Dead Reckoning method, providing the pedestrian heading regardless of the port of the sensor housing. The invention makes it possible to rely solely on the data of motion sensors of a housing BC carried by the pedestrian P and operating even in indoor or indoor, without any instrumentation of the environment. It should also be noted that the proposed dead reckoning technique makes it possible to complete an absolute location inaccurate or available from time to time by interpolating between absolute position measurements provided by other systems. It is then also possible, for example, to limit the calls to absolute positioning techniques and thus reduce the power consumption of positioning systems using the present invention. The present invention may therefore, for example, participate in a range of technologies called assisted GPS or "Assisted GPS" in English where the GPS location solution is assisted by instantaneous trajectory data. The present invention also constitutes an important complement to WIFI-type radio localization techniques, for example, which are imprecise in nature. One hypothesis is that the sensors are worn by a human being, and that his activity is here walking, one can expect a little more information induced by the walking pattern (or "walking"). ), and therefore specific movements and motion measurements are induced on the motion sensors of the housing. In other words, the movements printed on the sensor housing BC, due to the wearing of the sensor housing 10 BC by the pedestrian, and due to the walking or running activity of the pedestrian P, are not random and have remarkable characteristics. . These particular movements are printed on the sensor housing BC and are therefore measured by the motion sensors it embeds. In what follows, this hypothesis is considered to have been made (the sensors are worn by the pedestrian, who walks or runs). As a first consequence, and in accordance with the above, ie that the pedestrian is a walking (or short) human being, there is therefore a natural direct relationship between his body orientation in the frame of reference and the trajectory that he follows. As shown in Figure 3 or Figure 2, due to human morphology, it can be assumed that the heading of the trajectory is provided by the Anteroposterior AP direction of its body. Therefore, the problem to be solved in order to determine the unknown rotation operator QRT is to solve the problem for the unknown rotation operator which links the trajectory mark RT and the user's body mark or pedestrian mark Rp, because assumes that the orientation of the pedestrian marker provides the heading of the trajectory. The proposed solution works under this assumption (the heading of the trajectory is equal to or equivalent to the heading of the anteroposterior axis), which is not a limitation for a normal case of walking or running. The housing reference RB may be linked to the reference reference RR by the rotation transformation operator QBR, since it is considered that the sensor housing BC is able to provide its orientation in the reference reference RR by way of the conventional calculation. of the IMU attitude, known to those skilled in the art.
[0049] A pedestrian carrying a BC motion sensor housing and performing a walking or running activity is considered. The BC sensor housing can be carried by hand, for example for a smartphone or tablet, in consultation mode (typical use of a smartphone or tablet), or carried by hand in a pendulum, or placed in a pocket tied to the chest, in a trouser pocket, in a bag worn over the shoulder, or in a backpack. It can also be worn by hand in phone mode, so close to the ear. These positions are non-limiting examples. They remain valid for other electronic devices such as interactive glasses (then the port is linked to the user's head), electronic accessories. These different cases are illustrated in Figure 4. Without restricting the list of possible positions to the previous list, this list shows the extent of positions generally found. In addition, over time, the port of the BC sensor housing may vary. As patent FR2942388 discloses, walking activity generates movements of which certain characteristics are remarkable. A stride is the basic period of the walk (or race) reproduced substantially identically over time. Each stride includes the movement of the left foot and then the right foot (or in reverse order, according to convention). We will speak of stride to designate the base period of the march (or race) and of step to describe the period of the signal corresponding to the alternative pose of the right foot and the left foot. The rate of step is double the stride rate as shown in FIG. 5. The frequency generally observed for a walking activity is reduced to the frequency band of 0.5 to 2 Hz. The high frequency is higher if we want to include racing activities. It will therefore be common and useful to reduce the motion signals studied (for walking / running) to this frequency band, for example by a bandpass filter.
[0050] It is therefore interesting to note that the motion signals captured by a sensor substantially integral with the chest or pelvis of a pedestrian then have two peaks of power at two remarkable frequencies, one linked to the rate of stride, the another related to the movements of each leg, the pace of the pace. For non-pathological, symmetrical steps, these two frequencies are linked in a ratio of approximately two.
[0051] It is therefore known that the walking activity induces movements with remarkable characteristics. We propose to give a description in the pedestrian reference Rp as defined in Figure 3. Indeed, for example according to FR2942388, it is found that the movements of the thorax in translation along the axis Medio Lateral ML are essentially marked by movements at the frequency of the stride, that the movements of the thorax in the directions of the vertical axes VT and Anteroposterior AP are essentially marked movements at the rate of the steps. Thus, (i) the translational movement signals due to walking / running along the Medio Lateral axis ML essentially have a power peak at the rate of the stride (they have little power at the pitch frequency). (ii) The translational movement signals due to walking / running along the Vertical VT and Anteroposterior AP axes essentially show a peak power at the pace of the step (they have little power at the rate of stride). (iii) The peak power of the signals due to the translational movement along the Medio Lateral axis ML is at a frequency two times lower than the peak power of the signal of the translational movement along the axes AP-Anterio-Posterior or Vertical VT. Finally, note also that the translation movement signals along the axis VT and AP (both essentially comprising power at the pitch frequency) have a constant phase shift of an angle of about 7/2. It is thus found that there is information related to the orientation of the body, which is present on movement signals induced by the movements of the body of the pedestrian. Thus, when the motion signals sensed by the motion sensors are expressed in the housing reference RB, information is available relating to the orientation of the body relative to the sensor housing BC. Likewise, if the motion signals picked up by the motion sensors are expressed in another reference (for example the reference reference RR), then information relating to the orientation of the pedestrian's body relative to the reference mark is available. reference RR. So, more concretely, and to fix ideas using a canonical case, when a case equipped with a three-axis translational motion sensor (X, Y, Z), such as an accelerometer which is commonly used as sensor of the movements related to translations, is fixed or integral with the thorax or pelvis, and that the reference of the housing BC in which the sensor delivers its signals is aligned with the pedestrian mark Rp (ML, AP, VT), the signal X-axis motion sensor has all the outstanding properties of the motion signal along the Medio-Lateral axis ML (ie essentially a power peak at the frequency of the stride), the motion sensor signal according to the Y axis has all the remarkable properties of the motion signal along the AP Antero-Posterior axis (ie essentially a power peak at the pitch frequency), the motion sensor signal along the Z axis has all the remarkable properties the motion signal along the vertical axis VT (i.e. essentially a power peak at the pitch frequency). The movement signals delivered by the case BC thus have remarkable characteristics due to the alignment of the case BC with the pedestrian mark Rp (ML, AP, VT). Finally, the remarkable phase shift properties between movement signals in translation according to AP and VT are found on the signals of the motion sensors between Y and Z. For rotational movements, it would be appropriate, for example, to use a rotation sensor, such as a gyro. Similar to the translation characteristics explained above, when the mark of the sensor housing is identical to the pedestrian mark, all the remarkable characteristics of the pedestrian's rotational movements (the rotation signals according to FIG. ML axis essentially have power at the pitch frequency along the axis AP, they have essentially the power at the frequency of the stride, as listed above.
[0052] When the orientation of the sensor housing BC is aligned with the Plew mark Rp (therefore the path mark RT), we find is therefore necessarily these remarkable properties on each of the measurement axes. It is now possible to describe the basic principle of the invention. Indeed, according to the above, when this orientation is arbitrary, that is to say when the transformation operator QPB is not reduced to the identity as described above (the orientation of the sensor housing BC n ' is not "identical" to that of the pedestrian or trajectory), but unknown, it follows that a good QPB estimator must allow to find these remarkable properties, on the motion signals from the sensor housing BC and transformed by the QPB operator. Indeed, the raw pedestrian movement signals from the sensor housing BC, then present mixtures of remarkable properties (because they are each of the combinations of signals according to ML, AP, VT), but transformed by application of the QPB operator provides the signals in the pedestrian frame Rp (or RT trajectory) and these transforms must then again present the remarkable properties identified on the axes of the pedestrian mark Rp (ML, AP, VT). It is thus the object of the invention to exploit these remarkable properties in order to estimate the unknown operators QPB (or QRT). The concrete example of the case in which the motion signals are expressed in the housing reference RB proposes to exploit the remarkable characteristics of the movement signals of a pedestrian walk to estimate the unknown rotation operator QPB between the housing reference RB and the pedestrian mark R. Moreover, the same principle applies simply between the pedestrian mark Rp and any mark in which it is possible to express the movement signals. Thus, by knowing the orientation of the sensor housing BC in a reference reference RR (such as the Terrestrial reference), then it is possible to express the raw motion signals measured by the sensor housing in the reference reference RR and then d apply the same principle as before to estimate the rotation operator QRP OR QRT between reference reference RR and pedestrian reference R. We can then estimate this unknown operator QRT, as we have proposed to estimate the operator unknown QPB. Those skilled in the art know many methods that can implement in practice the principle of the invention. Indeed, our invention assumes that a good estimator of the unknown orientation (for example QPB) must transform the signals of the reference in which they are expressed (for example reference box or other reference mark, for example linked to the Earth) in such a way that they again (in whole or in part) display the remarkable characteristics of pedestrian movement signals in the pedestrian marker. Depending on the sophistication of the method, the available motion sensors, and the computing power available, it is possible to exploit some or all of the outstanding characteristics of pedestrian motion signals. Those skilled in the art can exploit methods of finding the parameters of the operator of transformation in rotation (three independent parameters) for example by technique of minimization (or maximization) of criteria built on the respect of remarkable characteristics.
[0053] It is indeed possible to construct criteria that are representative of the difference between the characteristics of the motion signals transformed by a candidate orientation operator and the remarkable characteristics of the movement signals in the pedestrian marker, and to take as the best estimator from the unknown orientation the one that generates the best criterion. In the case of a deviation criterion (the criterion is all the greater as the characteristics of the signals transformed by the candidate rotation operator do not respect the characteristics of the movements in the pedestrian marker), the best estimator will be that which generates the smallest gaps. Thus, for example, as long as the measured translational motion signals (which are optionally transformed from the signals from the housing into a selected reference frame) and transformed by the candidate operator and thus potentially representing the signals along the axis Anteroposterior AP of the pedestrian thus predicted by the candidate operator do not present a power essentially at the pace frequency, so the gap remains large and the candidate operator is not retained. In fact, the candidate operator is not the good one as long as we do not find this remarkable characteristic of the movement signals in translation along the Anteroposterior AP axis. It is possible to build search methods for the best candidate operator by iteration, for example, by proposing successive candidate operators and retaining the one with the best criteria. You can go through all the space of the transformation operators in rotation and choose the one that provides the best criterion as a solution. It is advantageous to use more efficient minimum or maximum search methods, such as gradient descents, for example. It is a question of traversing a landscape whose coordinates are the parameters of a transformation operator in rotation (ie three independent parameters), to calculate the value of the criterion of respect of the remarkable characteristics after having transformed the signals of movements in the reference of the pedestrian predicted by the candidate operator.
[0054] In addition to these automatic search methods by criterion optimization, it is also possible to analyze a particular criterion and to propose a direct analytical solution that provides in a single calculation the best estimator of the operator in unknown rotation. It is also possible to provide mixed solutions by combining direct analytical resolutions where possible and optimization methods when direct expression is not possible. For example, it is possible to design a QPB operator estimation method, which presents candidate operators (taken from all the possible transformation operators in rotation), the best candidate or candidates that is or is retained as a rotation estimator. QPB is one or more of those transforming the motion signals measured in the housing reference into signals which exhibit (in whole or in part) remarkable characteristics of the pedestrian marker movement signals Rp (ML, AP, VT). The best solution (the best candidate operator, the "winner") is then the one for which the correspondence between the properties of the signals transformed by this best solution and the remarkable properties of the walking / running signals in the pedestrian marker is the best. It is seen that it is potentially useful to combine several remarkable properties, so as to best estimate the unknown operator, limit indeterminacy, and provide an estimator most insensitive to noise. Note that some remarkable features may be more difficult to observe than others. They are not relevant to the criterion. Indeed, depending on the nature and the performance of the sensors, the type of step analyzed, some remarkable features will be more or less easy to identify on the signals, so it is appropriate for those skilled in the art to choose from the list, so to build an effective estimator.
[0055] It therefore appears that it is not at all obvious, as the prior art suggests that the direction of the path of the pedestrian is automatically given by the sole detection of the direction of a translation movement signal " powerful". Illustratively, we show that some high power signals occur along the medial-lateral axis ML of the pedestrian, perpendicular to the direction of movement of the pedestrian. Reference can be made to FIGS. 12 and 14. In one embodiment of the invention, these remarkable characteristics of the movements related to the chest or pelvis of the pedestrian are used: the signals of the movements in translation carried out by the thorax or the pelvis of the pedestrian during a walking activity (generalizable to the race) according to the vertical axis VT essentially have power at the step period ("step cadency" in the English language), - the signals of the movements in translation carried out by the thorax or pelvis of the pedestrian during a walking activity (generalizable to the race) along the Antero-Posterior axis AP present essentially power to the step period ("step 30 cadency" in English), and - the signals of the movements in translation carried out by the chest or pelvis of the pedestrian during a walking activity (generalizable to the race) according to the Medio-Lateral axis ML essentially present the a power at the stride frequency ("stride 35 cadency" in English), and - The frequency of stride is twice lower than the frequency of steps and corresponds to the frequency of stride On the other hand, it is also found that the phase shift 5 between the pedestrian translation movement signals along the Vertical VT and Anterior-Posterior AP axes (both having the same power characteristic essentially present at the pitch frequency) is close to 7/2 . This phase shift is therefore also a remarkable feature that we can introduce into the list of remarkable properties and thus into an estimator or a calculation sequence of the unknown rotation operator. Note the following very important element as well. Motion sensors do not have to be placed directly on the chest, thorax or pelvis of the pedestrian. It is only necessary that, by mechanical means, a sufficient part (measurable by sensors) of these motion signals be transmitted to the housing. Thus, it will be seen that the translational movements of the bust of a pedestrian are transmitted to a sensor housing carried by hand by the pedestrian in consultation mode for example. In this mode of carrying, the arm is a mechanical element that transmits the movements of the bust to the hand and these movements can then be measured. The head is also a position on which these properties are well transmitted. The case of pendulous limbs (pendulum arms for walking, for example), legs are exceptions and the remarkable properties of the movements of the pelvis should then be reviewed. When the pedestrian practices walking or running activity, the set of characteristics representative of movement signals in translation of the pedestrian bust represented in a pedestrian mark Rp = (AP, ML, VT) defined by the axes of the pedestrian Anteron -Posterior AP, Médio-Lateral ML, and Vertical VT includes: - the signal due to the movement in translation along the Medio-Lateral axis ML 35 has essentially power at the stride rate; the signal due to the translational movement along the anteroposterior axis AP has essentially power at the pace of the step; the signal of the movement in translation along the vertical axis VT essentially has power at the pace of the step; the signals due to movements in translation along the Vertical VT axis and along the Anterior-Posterior AP axis exhibit a substantially constant phase shift and close to 7/2; and the pace rate is substantially double the stride rate.
[0056] Moreover, other remarkable characteristics of the movement signals of the pedestrian bust P, this time in rotation can be exploited. Just as the accelerometer is a first choice simple and potentially low cost to capture the movements in translation of a pedestrian, a gyrometer is then for example a good choice of motion sensor rotation. Low cost gyrometers are used to measure instantaneous rotational speeds and are adapted to capture the rotational movements of the housing in which they are inserted. We then have the complementary list of remarkable properties on the rotational movements of the bust or thorax or pelvis following: the signal due to the rotational movement along the Medio-Lateral axis ML essentially has power at the rate of the step; the signal due to the rotational movement along the anteroposterior axis AP has essentially power at the rate of the stride; and the signal due to the movement in rotation along the vertical axis VT essentially presents power at the rate of the stride. Thus, all the remarkable characteristics of the movement signals of the pedestrian bust are larger than all the remarkable characteristics of the single translation movement signals. One can add a set of characteristics related to the rotational movements. One can therefore include in the calculation procedure of the unknown orientation operator a combination of the remarkable characteristics of translation and / or rotation, for example to improve the robustness of the estimator or its sensitivity to noise, to lift indeterminations. Note also that it is also possible to exploit a set of remarkable characteristics of the movements of a pedestrian who would be linked to a swinging arm, for example. If the motion sensor is rather sensitive to the movement of the arm, eg in the case of the port of the sensor in the hand and the arm being used as a pendulum when walking or running, then we can exploit the features that follow .
[0057] When the pedestrian is walking or running, the set of characteristics representative of pedestrian arm movement signals represented in a pedestrian marker (RP = (AP, ML, VT)) defined by the axes of the anterior pedestrian posterior AP, medio-lateral ML, and vertical VT, comprises: the signal due to the translational movement along the Antero-Posterior axis AP essentially presents power at the stride rate; the signal due to the movement in translation along the vertical axis VT essentially has power at the rate of the pitch; the signal due to the movement in translation along the axis ML essentially has power at the rate of the stride; the signal due to the rotational movement along the medio-lateral axis ML essentially has power at the rate of the stride; and the rotation signal due to the rotational movement along the vertical axis VT essentially presents power at the rate of the stride, this property being characteristic of a rocking movement of the arm. Other noteworthy features can be added to the above lists, to accommodate the specificities of the movements of other parts of a pedestrian's body. If the motion sensor is essentially marked by the movements of the bust or of a pedestrian member, depending on the nature of its measurement (for example rotation or translation), it is necessary to exploit all or part of the subset of the characteristics. remarkable correspondent.
[0058] The principle of the invention of exploiting the existence of remarkable characteristics of the movements of a pedestrian in a walking / running situation has been exposed. It is used to estimate the unknown rotation operator between the reference mark in which the motion signals are generated (from the sensors of a box carried by the pedestrian) and the pedestrian mark Rp. The pedestrian mark Rp being equivalent at the trajectory mark RT, the rotation operator estimated by the invention thus connects the reference mark in which the motion signals are generated, at the trajectory mark RT. The principle of estimation of the unknown operator consists in calculating, for a candidate operator, a criterion comparing the characteristics of the motion signals transformed by this candidate operator, and the remarkable characteristics of the pedestrian movements in the pedestrian axis Rp. set of candidate operators, the best estimator will be the one that presents the best criterion, ie which makes it possible to find the remarkable characteristics of the movements of a pedestrian in the pedestrian frame Rp. The computation can be carried out analytically and / or by a calculator which implements a criterion minimization method.
[0059] Moreover, all the remarkable characteristics of a pedestrian's movements can be composed of several elements and that, therefore, it was possible to introduce more information into our estimator. A greater amount of information makes it possible to better estimate the unknown rotation operator, with probably a higher calculation cost and potentially the need to integrate several types of sensors into the housing. It should also be noted that several housings distributed at different locations on the pedestrian's body can collaborate and consolidate the pedestrian trajectory orientation estimator. Indeed, a configuration that can commonly be considered is that the pedestrian is equipped on the one hand with his smartphone on the one hand and an additional accessory connected to the smartphone, such as interactive glasses and / or an accessory worn in the wrist, one and / or the other being equipped with motion sensors. Thus, the invention can be applied to both devices and deliver their estimated orientation of the path of the pedestrian, by the method of the invention. The principle applies to multiple devices. This multiple configuration is in itself interesting, because the redundancy of information makes it possible to build a better estimator of orientation by merging the multiple estimates. The merge method can be simple and for example combine the estimated multiples or estimates into an average. It is also possible to weight the estimated multiples, for example considering a priori their reliability according to their position on the body. Thus, for a position equipment a priori fixed on the pedestrian, such as the glasses placed on the head or wrist strap, the estimator can be considered to provide greater reliability than that of equipment which, a priori likely to change position on the body of the pedestrian, such as a smartphone. The weighting of the different estimators is then fixed a priori. Note also that the weighting can be dynamic, that is to say, not fixed in time. Indeed, as it is the object of the French patent application FR 1353616, filed, but not published by applying a detector for changing the position of the sensor housing on the pedestrian, for example by detecting variations in horizontality of the housing sensor, it is possible to weight the estimate dynamically over time. Indeed, during transitions, the trajectory orientation estimator is considered less reliable, since the estimator takes a certain time to converge. We propose here a particular embodiment interesting in many cases usually encountered in practice.
[0060] A rotation operator is completely determined by three parameters. In a previous paragraph we quoted as determining parameters an axis of rotation (thus a unit vector with two independent parameters) and an angle of rotation around this axis. This is how operators of transformation in rotation according to the formalism of Quaternions are represented. According to the equivalent formalism of the angles of Euler or Cardan, the operator of transformation in rotation is determined by three angles of rotation. He has three degrees of freedom. In all the many possible representations of a transform transformation operator, three independent parameters are needed to set the operator. It is then necessary to use a sufficient number of remarkable characteristics to correctly and uniquely determine these three degrees of freedom. The right set of remarkable features is not always easy to determine. To simplify this estimation, it is proposed to use a reference reference horizontal linked to the Earth, and the signals from the sensor housing BC are expressed in this reference reference RR, before estimating the rotation operator QRT. The reference mark then has a vertical axis, and two horizontal axes. Since the most common pedestrian marker is (ML, AP, VT) which also has a vertical axis and two horizontal axes, the unknown transformation operator between the reference mark and the pedestrian mark is limited to a vertical axis rotation. . The problem is then equivalent to a problem at a single unknown angle, instead of three angles or three parameters in the general case. This mode of implementation is adapted to a large number of situations commonly encountered, such as those of a smartphone user, interactive glasses or a user-worn accessory equipped with the sensors usually encountered in this type of device. . By definition of the reference linked to the human body as defined in FIG. 3, the vertical axis VT of the pedestrian is directed according to the vertical, that is to say along the axis of gravity. The Medio-Lateral axes ML and Antero-Posterior AP therefore form at each moment a horizontal plane. Thus, the problem to be solved in order to determine the unknown orientation between the reference mark in which the motion signals are expressed and the pedestrian mark Rp is to find the QRT rotation transformation operator which transforms the mark of the signals into the pedestrian mark. Rp. It is already easy to see that this transformation operator must already transform the plane of the signal marker into a horizontal plane, since the pedestrian marker has such a horizontal plane, which determines two of the three unknowns of the transformation operator in rotation. It is therefore convenient to immediately express the motion signals in a horizontal frame, by means of a rotation transformation operator provided or estimated using the motion sensors. The remaining unknown operator is then only a transformation operator in rotation along the single vertical axis (ie a heading angle) and it can then be more easily estimated by the principle of the invention. It will thus be possible to limit the remarkable characteristics to be exploited, to reduce the uncertainty of the estimator.
[0061] In a large number of situations, it is possible, from the sensors present in the case BC carried by the pedestrian, to estimate the first operator who transforms the frame mark RB into a horizontal reference mark and to express the signals therein. movement measured by the BC case sensors. Indeed inertial sensors commonly present in smartphones can calculate the orientation of the BC housing in the terrestrial reference RR. The last unknown is then related to the heading orientation of the body of the pedestrian. This is precisely one of the applications of the invention, exploiting the remarkable characteristics of the movements of the body, consisting in finding this last unknown angle. It is therefore advantageous to choose a reference reference linked to the Earth, in which the motion signals measured by the sensors of the case BC are expressed. The predominantly considered motion sensors (such as accelerometers, gyrometers, and magnetometers) provide three-dimensional measurement vectors. The sensors are rigidly connected to the BC sensor housing. It can therefore be considered that the sensors thus deliver their vector measurements in a same reference frame RB, linked to this case BC, which is noted RB. Those skilled in the art are aware of the methods which make it possible to correct any misalignments between the axes of the sensors themselves and thus which provide the measurements of the sensors in this same and unique reference box RB. As mentioned previously, techniques are also known for providing the orientation of the BC housing in the RR reference mark thus determined, fixed with respect to the Earth and having a horizontal plane and a given heading. It should indeed be noted that the sensors A, G, M, conventionally present in smartphones make it possible to constitute an attitude center (or "Inertial Measurement Unit" in English) which, for example by sensor data fusion technique , provides the orientation of the housing reference RB with respect to a reference reference RR linked to the Earth, that is to say the transformation operator in rotation QBR. Many combinations are possible between the sensors A, G, M in order to arrive at a satisfactory estimate of QBR.
[0062] The operator remaining to be estimated is then the operator QRT which is reduced to a rotation operator along the vertical axis VT. It is determined by a single angle, which greatly reduces the size of the space of the possible solutions. Indeed, even in the case of a poorly optimized search, it will suffice to traverse the space of the possible angles, this space being reduced to a single dimension, to form the corresponding rotation operator according to Z, to calculate the difference between remarkable features. The smallest gap is used to select the unknown angle. This reduces the size of the search space from three to one dimension.
[0063] We then present in detail a particular mode of implementation of the invention, which exploits this feature. The present invention is applicable to any location problem, whether indoors or outdoors. For some particular cases, for which one would look for a location solution in moving places with respect to the terrestrial landmark, such as a ship, the present invention would still apply, one skilled in the art would be able to apply an angle correction to reorient the map of the location moving in the reference landmark when needed.
[0064] The process input is a three-axis motion signal provided by a three-axis accelerometer, a three-axis magnetometer, or a three-axis gyrometer. The invention can be applied to each of these three sensors, or to a combination of these sensors. When looking for a motion signal related to the acceleration of a body, the best sensor to choose is an acceleration sensor. When looking for a motion signal related to the rotation of a body, it is possible to choose a gyrometer or a magnetometer. The method of the invention applies regardless of the choice made. In order to better evaluate the unknown angle of rotation, it is possible to envisage using a combination of sensors A, G, M. From the signals of the sensor housing BC, the orientation QBR of the sensor housing BC is estimated in the reference mark of terrestrial RR reference, for example by a method of inertial central type. The reference reference thus chosen is linked to the Earth. It has a horizontal plane formed by the XR and YR axes, the ZR axis being vertical. The unknown transformation operator QRT between such a reference reference RR and the trajectory mark RT (which is called equivalent to the pedestrian mark Rp and taken here equal to the pedestrian mark Rp) is then restricted to a rotation operator according to the axis vertical. Techniques known from the state of the art make it possible to estimate the orientation of the sensor housing BC in a reference reference RR from a combination of inertial sensors, such as accelerometers which easily provide information related to the angles of roll and pitch ("Roll" and "Pitch" in English) in the terrestrial reference, Gyrometers that provide the rotational speeds of the sensor housing, magnetometers that measure the Earth's magnetic field and allow to determine a cap of the sensor box relative to the North of the Earth. Depending on the class of sensors used, it will advantageously be possible to use single-gyrometer solutions, Accelerometer-Gyrometer or Accelerometer-Gyrometer-Magnetometer solutions. Note that the definition of the heading of the reference reference RR can be chosen according to convenience. The orientation estimator of the trajectory object of the invention will then provide the heading of the trajectory relative to the reference mark. Thus, a benchmark whose course is known in relation to the conventions of cards conventionally used for geographic tracking will be advantageous. Maps are typically located relative to the geographic North, so a benchmark with a similar convention will be advantageously used.
[0065] In the example of a smartphone, an accessory, and interactive glasses, the sensors present make it possible to estimate this orientation operator QBR of the housing in the terrestrial reference frame. The software computing solutions of the operator are often also embedded. As is well known to those skilled in the art, relying solely on an accelerometer provides a noisy estimate for this operator because the accelerometers provide the sum of the gravitational signals, which contain the signals useful for estimating roll and pitch angles, and the proper acceleration of the sensor housing, due to the trajectory of the sensor assembly, this acceleration component being considered a noise for the best estimate of roll and pitch angles. Furthermore, the heading angle of the reference mark can not be determined by a solution based solely on the accelerometer. However, when the movement is periodic, one can reduce the effect of the own acceleration by applying a low-pass filter on the accelerometer data and thus obtain a better estimate of the angles of roll and pitch. In any case, a solution based solely on an accelerometer does not provide any information on the yaw angle, which remains unknown. Other methods combining accelerometers and magnetometers offer the possibility of estimating the yaw angle, always to the detriment of the sensitivity to clean acceleration and therefore requiring some cleaning by previously mentioned filters. A solution based on a gyrometer is also possible. Those skilled in the art know the methods of calculating an orientation from a gyrometer signal that delivers rotational speeds. This sensor makes it possible, by an integration method, to find the orientation traveled since the first instant considered in the integration calculation. If the orientation is known at the first moment of integration, then we know at each moment the absolute orientation of the sensor housing. In addition to the need to know at a given instant absolute orientation, this method also has limitations related to the class of the gyro sensor. Indeed, a sensor with bias or sensitivity defects can generate significant errors at the output of the integration process. For example, we observe a drift that is all the faster as the bias is poorly understood. The current "consumer" class sensors exhibit drifts of the order of a few degrees per minute, which limits the processes based on the only gyrometers to very short usage scenarios. Better class sensors, which can be expected to be available at ever lower costs, show drifts of the order of a few degrees for scenarios of the order of the hour. When a high-performance gyrometer is available, it alone allows, according to its performance, to calculate the QBR operator at each moment, from an initial datum of the orientation of the box. There are thus central performance attitudes, independent of magnetic signals, strong accelerations, and little drifting for periods of time ranging from several tens of minutes to a few hours.
[0066] The best combination includes a combination of an accelerometer, a magnetometer and a gyrometer, which provides the complete orientation of the sensor housing in the terrestrial reference frame. Not only does it provide comprehensive orientation information, but it is not sensitive to either the proper acceleration or the effect of the angular drift of the gyrometer. Many techniques, such as those described in FR2934043, or FR2930335, or FR 1154915 can be applied to obtain the QBR operator, and this invention can therefore be applied. It is now considered that there is sufficient information on the rotation operator QBR connecting the sensor housing to the reference terrestrial reference. This operator is completely defined. As previously described, techniques known to those skilled in the art are capable of providing this QBR rotation operator. For the sake of clarity, it is considered in this part that the movement information of the body, are provided by the accelerometer of the sensor housing. As previously described, the following method can be applied to the gyrometer signals or to the magnetometer signals or to a combination of signals A, G, M. As mentioned above, and used in this mode, the motion signal can be expressed in a reference landmark by the application of the QBR operator and only an angle of rotation around the vertical axis is unknown and must be determined to move from the reference mark to the pedestrian mark (or trajectory since they are equivalent). It is therefore possible to use only a limited subset of pedestrian walking motion characteristics to provide the pedestrian's unknown heading angle in the land reference, which is the unknown we are looking for.
[0067] In what follows, we present a mode that only relies on a few characteristics of the movements in translation of the bust of the body of a pedestrian who practices a walking or running activity.
[0068] It should be noted in passing that when the pedestrian is at a standstill, it is no longer possible to estimate his heading, since the remarkable characteristics of pedestrian movements are no longer visible. This is not a problem since precisely the pedestrian does not move. The heading information then no longer makes sense.
[0069] From the three-axis measurement accelerometer signals 3A provided in the sensor housing, and from the knowledge of the QBR rotation transformation operator which provides the rotation between the reference of the sensor case RB and the terrestrial reference mark RR , it is easy to calculate the motion signals captured by the accelerometer in reference reference RR, one of whose axes is vertical called ZR (ie colinear to gravity), the other axes XR and YR being in the plane horizontal. A natural landmark is the landmark, North, East, Vertical. Since the ZR axis is the vertical axis, the YR axis can be considered as the North direction. The XR axis is selected to define an orthonormal basis (we say then pointing to the East). We have therefore defined our reference terrestrial reference, whose ZR axis is vertical, and the XR axis can be the North, the YR axis then being the East. It is then possible to provide the acceleration signals initially supplied in the reference RB of the sensor housing BC in this terrestrial reference, that is to say that AccX acceleration is obtained along the XR axis AccY acceleration along the YR axis, and AccV acceleration along the ZR axis which is equal to the vertical axis VT. In a preferred embodiment, this intermediate reference mark may be the reference terrestrial reference, defined by a "vertical" axis, a "North" axis and an "East" axis. The interesting step consists in obtaining motion information (in this case the acceleration signals) in a reference frame of which one of the axes is vertical, thus equal to the axis VT of the pedestrian marker R. It is then envisaged that the inventive steps which provide the remaining unknown angle which makes it possible to obtain the remaining rotation operator QRT from the reference reference RR to the pedestrian mark R. Thus, sensor signals which are supplied in the terrestrial frame, of which one of the axes is aligned with ZR gravity. Once again, it is considered that the pedestrian mark Rp and the trajectory mark RT are equal, and therefore that the QRT and QRP operators are equal. A rotation along a single axis connecting the pedestrian mark Rp to the reference landmark RR is now unknown, but we know that it is a rotation around the vertical axis with an unknown angle which we will note 0. This angle 0 is the Pedestrian heading angle in the land reference. Also, finding 0 solves the problem of trajectory heading, which is an object of the invention. The object of the invention is now to provide an estimate of this unknown angle θ, so that the rotation operator linking the reference landmark and the pedestrian marker is completely defined. As a reminder, the method according to the invention is divided into three main stages and uses the remarkable characteristics of the movement of the human body, as previously described. A complete calculation method, including the estimation of the rate of pedestrian walking, is presented. This information can be obtained in a different way from that presented here according to different modes known to those skilled in the art. The first step is to determine the step rate of walking. In this first step, it is, as an intermediate step, the delivery of the main frequency of the march (that is to say, to specify, the step frequency) of the walking or running activity. A preferred embodiment is to estimate the pitch frequency using the motion signal along the vertical axis and having the maximum power. Since the purpose here is to estimate the step frequency, any other method is adapted to the following steps as long as they provide the step frequency. The person skilled in the art knows many different methods for calculating the pitch frequency.
[0070] Calculating the pitch frequency from the motion signal of the vertical axis is a preferred embodiment. The step frequency can be estimated by another method, using another motion sensor, and / or using the acceleration standard, or using another detection axis, to make the estimate. the step frequency using signals in the sensor mark. As seen above, the exploitation of the vertical axis is advantageous, because when considering translation movements, one of the important characteristics listed is that the translation signal along the vertical axis has essentially power at the step frequency, which is what we are looking for in this first step. Since the walk activity provides a step frequency generally within a limited bandwidth, any method can be enhanced by high-pass filtering, low-pass filtering, or band-pass filtering filtering the signal. movement in the bandwidth of the pitch frequency. Typical values for walking activity are 1.0 Hz to 2.5 Hz. One skilled in the art knows that there are several techniques for estimating a pedestrian step frequency from motion signals. For example, it is possible to apply, for example, the technique of patent FR2942388. Techniques, time domain or frequency domain can be applied. The advantage of a method based on the vertical axis motion signal, which we describe as an example, is that the motion signal has the interesting property of having a maximum peak of energy centered around the frequency of motion. not and, for example, not be subject to the problem of mixing the pitch frequency and the stride frequency, which facilitates processing to obtain a reliable and accurate estimate of the pitch frequency. In this first step, once the pitch frequency has been estimated, the second substep of the first step is to provide the amplitude, energy or power of the motion signals along the XR axis and the YR axis (ie the AccX and AccY acceleration signals) for the pitch frequency. It can easily be understood that the purpose of this power estimate is to be able to match the remarkable features of the human gait, which is that along the anterior-posterior AP axis of the pedestrian mark Rp, the signal of a motion translation has a power peak at the pitch frequency, whereas on the Medio-Lateral axis, it shows no significant peak at this same frequency. For such a power estimate, it is possible to use, for example, the output of the Fourier transform of the AccX and AccY signals for the pitch frequency. Other techniques can be applied, in the time domain for example. One skilled in the art knows how to evaluate the amplitude, energy or power of AccX and AccY signals for pitch frequency. A narrow filter centered around the pitch rate frequency can be applied to the AccX and AccY signals, and the amplitude of the filtered signal can be calculated to provide the result.
[0071] In the many variants intended to estimate the signal strength on AccX and AccY at the pace of the step, it is also possible to apply a matched filter (or "matched filter" in English) on the AccX and AccY signals to estimate the signal strength at step frequency. This method is worth a short description because it is adapted to our case of a pedestrian. For this application, it is possible to choose as the impulse response of the matched filter the temporal signal according to AccV (that is to say according to the vertical). Indeed, it is known, as is listed in the list of outstanding characteristics of the movement signals of a walking pedestrian, that the acceleration signal according to the vertical component has substantially power at the frequency of the step. It is also known that the signal along the axis AP is essentially marked by power at the pitch frequency, and, interestingly, that this signal is out of phase with AccV by a constant value (approximately 7/2). Consequently, taking as the impulse response of a matched filter the acceleration signal in the vertical direction makes it possible, by applying this filter to the AccX and AccY signals, to extract from these two components the signal which is the best correlated with AccV. and thus to estimate the power, on AccX and AccY of the signal at the rate of the step.
[0072] Once the power of the signal at the frequency of the pitch extracted from AccX and AccY, the second step provides the angle of rotation remaining unknown between the two-dimensional reference or 2D (XR, YR), XR and YR being the horizontal axes of the reference marker and the Medio-Lateral axes ML and Antero- AP Posterior of the pedestrian mark Rp. This step is based on remarkable characteristics of the frequencies of human walking activity, which is that the movement along the Medio-Lateral axis ML does not have a step frequency (but a stride frequency signal), while the motion along the AP Antero-Posterior axis presents a pitch frequency signal (and not a stride frequency signal). Thus, the unknown angle makes it possible to transform AccX and AccY into the acceleration signals along the axes AP and ML. This transformation in rotation according to the vertical angle of 0 must therefore be such that, after transformation, all the power of acceleration signals at the gait rate is found only along the axis AP. The same logic could be applied as in Steps 1 and 2 by focusing on the frequency of the stride. We would then look first, from an estimate of the pace of the stride (or frequency of the stride) the power of signals according to AccX and AccY at this frequency. This frequency could simply be deduced from the frequency of the step by dividing the latter by a factor of 2, since the frequency of the stride is half of the pitch frequency. The power of the signal could then be sought at the frequency of the stride thus determined for example by a frequency transform technique. Then, from the power values of the AccX and AccY signals, find the angle that converts AccX and AccY into AP and ML, this time seeking that the peak power of the signal after transformation is found essentially along the ML axis. obtained with the candidate angle, since it is known that the power of the acceleration signal at the frequency of the stride is found essentially along the axis ML of a pedestrian. As noted, this second step provides the unknown heading angle of rotation modulo 7. So we get the direction of the axis AP, or direction of walking of the pedestrian. At the end of the second stage, we can not find the meaning of the Antero-Posterior AP axis, which is positive or negative, which means that we do not know the direction in which the pedestrian is moving. Thus one can not determine if the pedestrian is moving in one direction or the opposite, leaving two possible solutions. With this single remarkable characteristic of the movement of walking, there is always a solution of course under-determined at 7 meadows. With this single remarkable characteristic of the human movement movement, there is always an under-determined solution.
[0073] The third step provides the selection of the direction and raises the indeterminacy of modulo 7, from the passage of the marker (ML, AP) to the mark (XR, YR) from the analysis of the remarkable characteristic of phase shift between the signals of the accelerometer according to the axes Antero-Posterieur AP and vertical VT which is 7/2. The overall block diagram of one embodiment of the invention is shown in FIG.
[0074] Note that one can easily modify the solution presented here in detail to take advantage of other remarkable features of human walking, such as the presence of a power peak on the Medio-Lateral axis ML at the frequency of the stride and no power peak at the stride frequency on the AP Antero-Posterior axis.
[0075] This is a direct extension of the detailed solution presented here. It is also possible to use the same characteristics, in a different order or by methods equivalent to that described. FIG. 6 shows in block diagram mode a calculation method which implements an embodiment of the invention. (AccX, AccY, AccV) are the 3D motion signals delivered in the reference frame (XR, YR, ZR), ZR being vertical, and (XR, YR) forming a horizontal plane. The reference mark can be a landmark. The first step of this embodiment (i) calculates the step frequency (indicated by the index idx) of the walking activity from AccV. This first step also delivers the magnitude of the vertical movement AccV around the pace of the step, ie DFTv (idx) (it is here concretely a complex number with a module and a phase). In this embodiment, a Fourier transform is used, the pitch frequency is estimated from the frequency domain signal AccV. Then (ii) calculate the magnitude of the AccX and AccY signals around the pitch frequency given by the idx index. In this embodiment, these magnitudes DFTx (idx), DFTy (idx) are calculated by a Fourier transform around the frequency of the pitch given by the index idx. In a second step of this embodiment, the remarkable characteristic of the motion signals along axes AP and ML is exploited (the first, AP, has a majority power around the pitch frequency, the second a majority power at the frequency of the stride). In this second step, the values DFTx and DFTy are used to calculate the unknown rotation angle between the vertical axis that transforms the reference mark in the pedestrian mark. At the end of this step the unknown angle is determined to modulus Pi near. We have the direction of the axis of the trajectory, but not the direction. The third step takes into consideration the remarkable property that says that the signals according to AP and VT has a constant known phase shift (close to 7/2). This third step thus makes it possible to determine the unique angle and resolves the indeterminacy to Pi near. This step thus finally provides the desired angle that determines the transformation between the reference mark and the pedestrian mark, which determines the direction of the pedestrian in the reference mark.
[0076] In the remainder of the description, implementation details of the solution presented above are presented, according to the block diagram of the treatments presented in FIG.
[0077] Here follows the detailed description of the first step (frequency analysis). As previously presented, we focus on the characteristic of the translation signals (measured here with an accelerometer) which, according to the vertical axes VT and AnteroPosterieur AP, essentially of the power at the frequency of the pitch, whereas this frequency is absent according to the Medio-Lateral axis ML. We present a detailed mode of obtaining the pitch frequency from the acceleration signal along the vertical axis VT, and we present the calculation of the signal power along the horizontal axes AccX and AccY. The main frequency of the motion signals along the vertical VT (Rz) axes is first calculated. To calculate this frequency, the DFT (Discrete Fourier Transform) can be calculated on the motion signals along the vertical axis VT in the frequency range corresponding to the frequency range of the step (for example 1 Hz-2.5 Hz). . An index corresponding to the power frequency (or maximum amplitude) is obtained. Then, for this index, one calculates the DFT acceleration signals AccX and AccY, which represent the amplitude at the running frequency (here that of the step). To calculate these discrete Fourier Transforms or DFTs, one must first choose the size of the time window and the sampling frequency of the signals. Those skilled in the art can select the size of the time window, the sampling frequency of the signals, the precise calculating mode. However, by way of example, we present an applied process, which will enable those skilled in the art to apply variants with a solid basis of comparison. To do this, the maximum frequency considered is 2.5 Hz, and the Shannon theorem gives a minimum sampling frequency of 5 Hz. In practice, a sampling frequency higher than this frequency is chosen, for example a period of 20Hz sampling. Fe = 20 Hz (1) The minimum frequency that is to be distinguished is half of the minimum pitch frequency (0.5 Hz), corresponding to the minimum frequency of the acceleration signal along the Medio-Lateral axis ML . Thus, the time window used for the frequency analysis must contain at least one signal period of this minimum frequency. Window = 2s (2) The Fourier transform calculations are used to first extract the value of the pitch frequency from the acceleration signal along the vertical axis, and then, secondly, the power ( or amplitude) acceleration signals along the horizontal axes for this pitch frequency. Note that the proposed method is one of many possible methods. The illustration implementing a Fourier transform is particularly simple to understand because it is an operator known to those skilled in the art. Other techniques, for example by AutoRégréssif model (AR) of the signal, or by adapted filtering makes it possible to achieve the same ends. The objective is to extract the two powers of the horizontal acceleration signals at the pitch frequency, in order then to apply the remarkable characteristic of the movements in translation of the bust, thorax, or pelvis of a pedestrian, according to the Anteroid axes. -Posterior AP and MedioLateral ML. Fourier or DFT transforms do not need to be calculated for all frequencies, which considerably reduces the calculations.
[0078] We present here a method that recursively calculates a particular element of the DFT, which is given by the following formula: DFT = rk1: (1) xke2 ÷ 1ijk (3) 35 in which: xk represents the samples of the signal to analyze; n represents the number of samples in a time window (in this case 40); and j represents a discrete frequency index defined by: Fi = Fe L (4) The direct calculation of a DFT element requires n-1 complex products and n complex additions. In addition, the DFT is computed on a sliding window, proposing the recursive calculation of DFT: DF7) (t + 1) = (DFTi (t) -x (t-n-1)) e-2Zii (n- 1) + x (t) e-rin (5) This recursive version of the DFT requires two complex products and two complex sums. This is more optimized than the FFT algorithm when the computation is done on a sliding window and for a small number of frequency indices. On the Vertical VT axis, DFTj for j is to be calculated from 2 to 5, corresponding to the frequency range from 1 Hz to 2.5 Hz (equation (4)). Then, we find the maximum of these 4 DFT values to have the index jw 20 corresponding to the operating frequency. For X and Y axes, only DFTjW should be calculated. Two ways of calculating this DFT are proposed: 1. Calculate recursively DFTj for j = 2, ..., 5 by means of equation (5) then select the value corresponding to jw. 2. Compute recursively DFTjW using equation (5) each when jw (t + 1) = jw (t) or calculate DFTjW using equation (3) when jw (t + 1) # jw (t ). 30 The choice of the method depends on the variation of jw, if jw varies a lot (more than once a second), it is necessary to choose the first method of calculation, if not the second one. In this case, we choose the second. In a preferred embodiment of the second step of calculating the unknown angle of rotation, it is used that once the powers or amplitudes of the AccX and AccY acceleration signals are calculated along the two horizontal axes, at the step frequency, ie DFTx (idx) and DFTy (idx), it is necessary to apply the principle of the invention so as to find the angle by which it is necessary to operate a rotation of the axis mark XR , YR so as to find the pedestrian mark defined by the axes ML and AP in which the remarkable characteristic is expressed.
[0079] To express the unknown angle of rotation, the next step of the method consists in calculating the angle of rotation between the 2D mark formed by the XR and YR axes and the mark formed by the ML and AP axes. These two marks are represented in FIG.
[0080] Equation (6) provides the rotation in vertical axis of amplitude values of acceleration signals from the reference frame in which the estimate of the amplitude in the pedestrian marker is obtained. We can say that the rotation matrix thus formed with the angle 0 represents the candidate operator. If the candidate operator is good then: (DFTmi, [cos (0) -sin (0) 1 (1) FT.x (6) DFTAp) [sin (0) cos (0) iY) FTy) In fact, to apply the remarkable characteristic due to the walking movement of the pedestrian, we must find, by the transformation (6) applied to the doublet (DFTx (idx), DFTy (idx)), also noted (DFTxj, '' DFTyi,) with j , corresponding to the step frequency, that DFTAP is maximum, while DFTML is minimum. In the detailed example, it is proposed to look for the angle of rotation that maximizes the value of DFTAP. It is then shown that the search for this angle is possible with direct equations, so it is not necessary to deploy iterative search methods of the angle. Also, we look for the rotation that maximizes the DFTApr module, which is equivalent to solve the following optimization problem: max F (0) = max IlDFTxj, sin (0) + DFTyi, cos (0) 112 (7) It is possible to find an analytical solution to equation (7). Other methods are nevertheless possible. We present a method of calculation.
[0081] We can take the following notations: = a1 + = a2 + ib2 (8) We insert equation (8) in equation (7), and we obtain: F (6) = (a1 sin (0) + a2 cos (0)) 2 + (b1 sin (0) + b2 cos (0)) 2 (9) Solving equation (7) is equivalent to finding the solution of the following equation: F '(61) = Acos (0) sin (0) + B (cos2 (6) - sin2 (0)) = 0 (10) wherein: A = ct + bi2_ - b2, B = a1 a2 + b1b2 (11) The equation (10) ) is equivalent to: -2Btan2 (0) + 2 tan (0) - -2B = 0 AA (12) This equation has two solutions for tan (8) defined in equation (13) corresponding to the minimum and maximum of the equation (9): = - A 1 + _ 11 + 4B) -A2, S2 = - 2B 1 -, 11 + 4B2 (13) Also, the two possible solutions of equation (7) are: 01 = atan (S1), or 02 = atan (S2) (14) To find the right solution, we evaluate DFTApr, for the two solutions and find the one that maximizes DFTAp In addition to the market cap, we can calculate a value of "confidence" of the course calculation thus determined. Indeed, to calculate the heading, we maximize the power at the pitch frequency along the axis AP. Residual power along the ML axis can also be calculated. If the difference between these two values is large, i.e. above a threshold, we trust the calculated heading. t (IDFTApl-IDFTmLI) (15) WHconfidence IDFTApi Thus, we build a value between 0 and 1 that tends to 1 if all the power is found on the AP axis and that tends to 0 if the powers are distributed between AP axis and the ML axis. FIG. 8 shows the time signals of the unknown angle as determined by an implementation mode of the invention, as obtained from step 2. The value of the unknown angle is represented by 0. The truth, estimated or otherwise provided, is represented by a writ. We present two intermediate values of the calculation of the unknown angle by 01 and 02. These two angles are solutions which cancel the derivative of the criterion which must be maximized. The solution that maximizes the criterion and thus comes from step 2 is represented by Omax. At the end of this step, the angle is determined modulo 7 pres. Figure 8 shows the results of the calculation of the angle of rotation, and illustrates the two solutions 81 and 82, - e max, and Bref the actual angle. We can see that the chosen solution is the right one and is close to the real angle. At the end of this step, the heading angle is determined modulo 7 pres. We know the direction of the pedestrian's walk, that is to say the axis along which it moves, but we can not give the direction of its march in this direction. In a preferred embodiment of the third step, this indetermination is thrown. At the end of the preceding step, which uses the remarkable characteristic of the movements in translation of the bust at the pitch frequency, the angle of heading to modulo 7 is estimated. There remains therefore an indeterminacy on the direction of the pedestrian mark Rp with respect to the reference reference RR (i.e. between forward and reverse). To remove this indeterminacy, it is possible to introduce an additional characteristic, linking the phase difference (or delay) of the acceleration signals along the vertical axis VT and along the anteroposterior axis AP. According to this remarkable characteristic, due to the movement of a pedestrian in a walking or running situation, the phase difference must be close to 7 / 2.35 The final heading angle is equal to the angle calculated in the previous step e modulo 7 m . To overcome this uncertainty, the phase shift between the DFT of the VT axis and the AP axis is calculated. (I) = angle (DFTAp jw) - angle (DFTv jw) (16) If the calculated phase shift is close to 7/2, do not apply the correction to the angle, if it is close to 37 / 2, a correction of 7 is applied to the angle. In practice, we compare cp to 7 to make the decision. FIG. 9 shows the time signals of the unknown angle as determined by an implementation mode of the invention, as obtained from step 3. This step makes it possible to remove the indeterminacy of modulo Pi existing after the Step 2. The remarkable property exploited is the phase shift angle between the motion signal according to AP and VT. This phase shift is theoretically close to Pi / 2. Fig. 10 shows an angle as obtained from step 2 (before correction) which has a deviation of Pi from the actual angle. The phase criterion compares the phase shift between the VT and AP axes. If this phase shift is greater than a threshold (here fixed at Pi) then a correction of Pi is made on the angle before correction. In the examples which follow, the position of the sensor considered in the experiments, the results are presented for which the orientation of the sensor housing related to the position of the trajectory is known: - hand, in consultation mode, - portrait orientation, for where the orientation of the sensor is approximately equal to the course of the trajectory, - landscape orientation, for which the orientation of the sensor has a constant deviation of 90 ° with the course of the trajectory The trajectory used for the experiment is a path Rectangular as shown in Figure 10 which has a rectangular shape course, with 4 segments AB, BC, CD, DA respective heading 17 °, 107 °, -163 °, -73 °. The starting point belongs to segment AB and is marked by a solid circle. For the first test, the user walks with his smartphone in 5 orientation portrait. Also, the user walks in the direction of the heading of the trajectory. As illustrated in FIG. 11, the pedestrian heading calculated using the method of the invention and the actual heading are very close. The delay is about one second, which is half the slippery window used for calculating the discrete Fourier transform. As long as the user is in walk mode, the result is good. In other words, FIG. 11 shows different time signals illustrating the result as a result of a particular embodiment of the invention applied to the path shown in FIG. 11. The first graph shows the vertical acceleration component. The second graph shows (i) the angle θ of the trajectory as estimated by a particular embodiment of the invention (ii) as well as the heading angle of the sensor housing, denoted Cap. In this example, the sensor housing is worn so that the heading angle of the sensor housing coincides with the course of the trajectory. This value is therefore the angle value with which the trajectory angle estimated by the invention must be compared. The third graph shows the difference between the angle Cap and the angle O. Figure 12 shows for illustration some periods of the acceleration signal, estimated in the pedestrian axis. Thus, the remarkable characteristics of the translational movements of a pedestrian are observed experimentally, in particular the acceleration signal along the axis ML essentially having power at the frequency of the stride, the acceleration signal according to FIG. AP axis has essentially the power at the pitch frequency, the acceleration signal along the VT axis has substantially the power at the pitch frequency and has a constant phase shift of about 7/2 compared to the signal according to AP .
[0082] Figurel2 shows the decomposition of the acceleration signals in the reference of the human body. The frequencies along the axes VT and AP are identical and two times lower along the axis ML.
[0083] In the following test, the user holds his smartphone in landscape orientation. The smartphone is heading to the right. Thus, the offset between the smartphone trajectory and the heading of the trajectory provided by the invention must be 90 degrees, which is correct, as illustrated in FIGS. 13 and 14. A result is presented according to the same formalism as FIG. 11, for this time a realization of the course where the sensor is held in "landscape" mode. The Cape angle is no longer confused with the Cape of the trajectory but spread about 7/2. The angle θ as estimated by the invention correctly estimates the heading of the trajectory. Figure 14 shows, in the same formalism as Figure 12, some periods of the acceleration signal in the pedestrian marker. The steps of the method described above can be performed by one or more programmable processors executing a computer program for performing the functions of the invention by acting on input data and generating output data in the device according to the present invention. 'invention. A computer program can be written in any programming language, such as compiled or interpreted languages, and the computer program can be deployed in any form, including as a standalone program or as a subprogram or function, or any other form suitable for use in a computing environment.
[0084] A computer program can be deployed to run on a computer or multiple computers on a single site or on multiple sites distributed and interconnected by a communication network.

权利要求:
Claims (19)
[0001]
REVENDICATIONS1. Method for determining the orientation of the path followed by a pedestrian (P), associated with a trajectory mark (RT), relative to a reference mark (RR), said pedestrian being provided with a sensor housing (BC) comprising a sensor assembly (EC) comprising at least one motion sensor, comprising the steps of: generating data representative of the movement of the sensor housing (BC) from said sensor assembly in the reference frame (RR), and calculating the value of a first rotation transformation operator (QRT) representative of the reference reference (RR) orientation relative to the trajectory reference (RT), so that the data representative of the movement thus obtained in the previous step in the reference frame (RR), and transformed by said first operator (QRT), have at least one characteristic of a set of characteristics representative of movement signals of a n pedestrian, expressed in the pedestrian marker.
[0002]
The method of claim 1, wherein the reference mark (RR) is a terrestrial mark, and said generation of data representative of the motion of the pickup box (BC) is obtained from said pickup set in the reference mark (RR ) by applying a second rotation transformation operator (QBR), so as to determine the orientation of the trajectory marker (RT) in said terrestrial reference (RR).
[0003]
The method according to claim 2, wherein the reference mark (RR) and the trajectory mark (RT) comprise a common axis, so that the first rotation transformation operator (QRT) is reduced to an operator of transformation into rotation along the common axis.
[0004]
4. The method of claim 3, wherein the common axis is oriented in the direction of the Earth's gravity, so that the first operator of transformation in rotation (QRT) is reduced to a transformation operator in rotation along an axis direction of gravity Terrestrial.
[0005]
5. Method according to one of claims 2 to 4, wherein determining the orientation of the sensor housing (BC) in the pedestrian frame by composition of the second operator (QBR) with the first operator (QRT).
[0006]
6. Method according to one of claims 2 to 5, wherein there is provided a central attitude function providing the value of the second operator (QBR) of the sensor housing (BC) in the reference frame (RR).
[0007]
The method of claim 6, wherein the central attitude function calculates the second operator (QBR) from a combination of data provided by accelerometric and / or gyrometric and / or magnetic inertial motion sensors present in said sensor housing (BC).
[0008]
8. Method according to one of the preceding claims, wherein, when the pedestrian practice walking or running, said set of characteristics representative of movement signals of the bust, thorax or pedestrian pelvis represented in a pedestrian marker (Rp = (AP, ML, VT)) defined by the axes of the pedestrian anteroposterior (AP), mediolateral (ML), and vertical (VT), said pedestrian mark being linked to the trajectory mark (Rr), comprises the following characteristics: the signal due to the movement in translation along the medio-lateral axis (ML) has essentially power at the stride rate; the signal due to the movement in translation along the anteroposterior axis (AP) has essentially power at the pace of the step; the signal due to the translational movement along the vertical axis (VT) essentially has power at the pace of the step; the signals due to the movements in translation along the vertical axis (VT) and along the anteroposterior axis (AP), at the pitch frequency, have a substantially constant phase shift; the step rate is substantially double the stride rate; the signal due to the rotational movement along the medio-lateral axis (ML) has essentially power at the pace of the step; the signal due to the rotational movement along the anteroposterior (AP) axis has essentially power at the rate of the stride; and the signal due to the rotational movement along the vertical axis (VT) essentially has power at the rate of the stride.
[0009]
9. Method according to one of the preceding claims, wherein, when the pedestrian practices a walking or running activity, said set of characteristics representative of movement signals of a free member of the pedestrian represented in a pedestrian marker (Rp = (AP, ML, VT)) defined by the axes of the anteroposterior pedestrian (AP), medio-lateral (ML), and vertical (VT), said pedestrian mark being linked to the trajectory mark (RT) comprises the following characteristics: the signal due to the movement in translation along the Antero-Posterior (AP) axis has essentially power at the stride rate; the signal due to the translational movement along the vertical axis (VT) essentially has power at the pace of the step; the signal due to the rotational movement along the mediolateral axis (ML) has essentially power at the rate of the stride; the rotation signal due to the rotational movement along the vertical axis (VT) essentially has power at the rate of the stride.
[0010]
The method according to claim 8 or 9, wherein said one or more features are selected from the set of pedestrian bust movement characteristics of claim 8 or from the set of characteristics of the movements of a free limb. pedestrian according to claim 9, from an indicator characterizing the nature of the mechanical connection between the sensor housing (BC) and the pedestrian.
[0011]
The method according to one of claims 8 to 10, wherein said data representative of the motion of the sensor housing (BC) in the reference frame (RR) are generated from at least one accelerometer with at least 2 axes of measurement, and for which at least one of said one or more characteristics is that the acceleration signal due to walking / running along the main direction of the trajectory or along the anteroposterior axis (AP) essentially has a peak power at the rate of step.
[0012]
12. Method according to one of claims 8 to 10, wherein said data representative of the movement of the sensor housing (BC) in the reference frame are generated from at least one accelerometer with at least 2 measurement axes, and for which at least one of said one or more characteristics is that the acceleration signal due to walking / running in the direction perpendicular and horizontal to the main direction of the trajectory or along the mediolateral axis (ML) presents essentially a peak of power at the pace of the stride.
[0013]
13. A method according to claim 11 or 12, wherein the vertically rotational transformation operator (VT) is further determined such that the phase shift between the acceleration due to walking / running measured according to the vertical axis (VT) and the acceleration due to the step / stroke at the pitch frequency and transformed by said operator along the axis AP is between 0 and Tr, and is particularly 7/2. 20
[0014]
14. The method according to claim 11, wherein the operator for transforming in rotation along the vertical axis (VT) is determined from the amplitudes, at the pitch frequency, of the two horizontal components of the acceleration signal in the reference mark (RR).
[0015]
15. The method according to claim 12, wherein the vertical rotation transforming operator (VT) is determined from the amplitudes, at the stride frequency, of the two horizontal components of the acceleration signal. the reference mark (RR).
[0016]
The method of claim 14 or 15, wherein the pitch or stride frequency is determined from the acceleration signal along the vertical axis (VT).
[0017]
17. The method of claim 14 or 15, wherein the amplitude of the pitch frequency acceleration signal is determined by matched filtering of the acceleration signal in the reference frame (RR), according to the signal filter. vertical acceleration.
[0018]
The method of claim 9 or 10, wherein said data representative of the motion of the sensor housing (BC) in the reference frame (RR) is generated from at least one gyrometer to at least two measurement axes, and wherein at least one of said one or more characteristics is that the rotational speed signal due to walking / running along the mid-lateral axis (ML) essentially has a peak power at the rate of the stride.
[0019]
19. A method according to claim 9 or 10, wherein when the movements printed in said sensor assembly (EC) are essentially due to the movement of the chest-thorax-pelvis unit (P), at least the characteristic according to which the movement signal in translation at the frequency of the pitch has substantially power along the anteroposterior axis (AP), and when the movements printed in said sensor assembly (EC) are essentially due to the movement of a free member of the pedestrian (P), at least the characteristic that the signal of movement in rotation at the frequency of the stride has essentially power according to lateral medio (ML).

类似技术:

---

公开号 | 公开日 | 专利标题

---

FR3015072A1|2015-06-19|METHOD OF DETERMINING THE ORIENTATION OF A MOBILE TERMINAL-RELATED SENSOR MARK WITH SENSOR ASSEMBLY USED BY A USER AND COMPRISING AT LEAST ONE MOTION-MOVING MOTION SENSOR

---

EP1984696B1|2009-07-08|Motion capture device and associated method

---

EP2646776B1|2017-09-06|Method and system for determining the values of parameters representative of a movement of at least two limbs of an entity represented in the form of an articulated line

---

EP3185110B1|2020-09-02|Hybrid mobile element, method and device for interfacing a plurality of hybrid mobile elements with a computer system, and assembly for virtual or augmented reality system

---

KR20150108402A|2015-09-25|Method and apparatus for determination of misalignment between device and pedestrian

---

US9752879B2|2017-09-05|System and method for estimating heading misalignment

---

EP2350565B1|2014-06-18|Device and method for determining a characteristic of a path formed by consecutive positions of a triaxial accelerometer rigidly connected to a mobile element

---

FR2886501A1|2006-12-01|METHOD AND DEVICE FOR LOCALIZING A TERMINAL IN A WIRELESS LOCAL NETWORK

---

EP3359915B1|2019-12-25|Method for estimating the movement of a pedestrian

---

FR3012597B1|2019-08-09|INTERNAL AND EXTERNAL LOCALIZATION METHOD AND PORTABLE DEVICE IMPLEMENTING SAID METHOD

---

FR2959112A1|2011-10-28|SYSTEM FOR ANALYZING CROWDS OF A USER

---

Ren et al.2012|Movement pattern recognition assisted map matching for pedestrian/wheelchair navigation

---

EP3227705A1|2017-10-11|Electronic device for the near locating of a terrestrial object, and method of locating such an object

---

EP1788357A1|2007-05-23|System for locating pedestrian user

---

Chang et al.2013|Techniques for 3D Misalignment Calculation for Portable Devices in Cycling Applications

---

FR2939191A1|2010-06-04|DEVICE AND METHOD FOR ASSISTING THE GUIDANCE OF A PIECE IN DIFFICULTY OF VISUAL PERCEPTION, SUCH AS AN INDEPENDENT PERSON

---

FR2963422A1|2012-02-03|Method for estimating movement of person in building of city for indoor location of person, involves obtaining series of angular speed measurements and calculating frequency of steps completed by person from measurements

---

WO2019016473A1|2019-01-24|Method for estimating the movement of an object moving in a magnetic field environment

---

EP3211370A1|2017-08-30|Method for filtering signals from a sensor assembly comprising at least one sensor for measuring a vector physical field that is substantially constant in time and space in a frame of reference

---

CH703020A2|2011-10-31|Perennial 3D localization system of an individual progressing inside a building.

---

EP3559598A1|2019-10-30|Method for autonomously geolocating a person traveling on foot or by means of a non-motorized vehicle and associated device

---

FR2985588A1|2013-07-12|Method for locating mobile e.g. pedestrian, in e.g. building interior, involves estimating trajectory of mobile and position of vertical edges in built medium, and estimating errors and plane of flat vertical walls visible from trajectory

---

Asmar et al.2009|Vision SLAM maps: Towards richer content

---

FR2939517A1|2010-06-11|Stationary non-interactive artificial object geo-locating method for aircraft, involves acquiring position-object distance pair, and calculating geo-location of stationary non-interactive object from three position-object distance pairs

同族专利:

---

公开号 | 公开日

---

US11002547B2|2021-05-11|

---

WO2015091402A1|2015-06-25|

---

FR3015072B1|2017-03-17|

---

US20160313126A1|2016-10-27|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

US20120072166A1|2010-09-22|2012-03-22|Invensense, Inc.|Deduced reckoning navigation without a constraint relationship between orientation of a sensor platform and a direction of travel of an object|

---

US20120296603A1|2011-05-16|2012-11-22|Qualcomm Incorporated|Sensor orientation measurement with respect to pedestrian motion direction|

---

WO2013022714A1|2011-08-11|2013-02-14|Qualcomm Incorporated|Numerically stable computation of heading without a reference axis|

---

FR1154915A|1956-07-05|1958-04-18|Chempatents|Process for preparing para-diisopropylbenzene|

---

FR1353616A|1963-01-17|1964-02-28|Snecma|Process of execution of volute for centrifugal pump|

---

US6122960A|1995-12-12|2000-09-26|Acceleron Technologies, Llc.|System and method for measuring movement of objects|

---

US5744953A|1996-08-29|1998-04-28|Ascension Technology Corporation|Magnetic motion tracker with transmitter placed on tracked object|

---

US20070032748A1|2005-07-28|2007-02-08|608442 Bc Ltd.|System for detecting and analyzing body motion|

---

FR2896896B1|2006-02-02|2009-09-25|Commissariat Energie Atomique|METHOD FOR CLASSIFYING EVENTS OR STATEMENTS IN TWO STEPS|

---

FR2926971B1|2008-02-04|2011-02-25|Commissariat Energie Atomique|MARKET ANALYSIS DEVICE|

---

FR2930335B1|2008-04-18|2010-08-13|Movea Sa|SYSTEM AND METHOD FOR DETERMINING PARAMETERS REPRESENTATIVE OF THE ORIENTATION OF A MOVING SOLID SUBJECTED TO TWO VECTOR FIELDS.|

---

FR2934043B1|2008-07-18|2011-04-29|Movea Sa|IMPROVED ESTIMATING METHOD OF OBJECT ORIENTATION AND ATTITUDE CENTER USING SUCH A METHOD|

---

FR2942388B1|2009-02-26|2012-10-12|Movea|SYSTEM AND METHOD FOR DETECTING THE MARKET OF A PERSON|

---

FR2951825B1|2009-10-26|2012-04-20|Movea|SYSTEM AND METHOD FOR COUNTING DIRECTION CHANGES IN A PERSON|

---

US9357948B2|2010-12-01|2016-06-07|Commissariat A L'energie Atomique Et Aux Energies Alternatives|Method and system for determining the values of parameters representative of a movement of at least two limbs of an entity represented in the form of an articulated line|

---

US9366749B2|2011-04-15|2016-06-14|Qualcomm Incorporated|Device position estimates from motion and ambient light classifiers|

---

US20130023798A1|2011-07-20|2013-01-24|Intel-Ge Care Innovations Llc|Method for body-worn sensor based prospective evaluation of falls risk in community-dwelling elderly adults|

---

EP2740256B1|2011-08-04|2017-12-13|Google LLC|Moving direction determination with noisy signals from inertial navigation systems on mobile devices|

---

CN102997913B|2011-09-14|2015-11-25|意法半导体投资有限公司|For determining method and the device of gestures of object|

---

US20140357305A1|2013-05-30|2014-12-04|Indooratlas Oy|Orientation of mobile device measuring earth's magnetic field indoors|

---

US20150316579A1|2014-05-02|2015-11-05|Qualcomm Incorporated|Motion direction determination and application|WO2015039215A1|2013-09-17|2015-03-26|Invensense, Inc.|Method and apparatus for determination of misalignment between device and pedestrian using vision|

---

EP3194890B1|2014-09-21|2020-02-26|Athlete Architect LLC|Methods and apparatus for power expenditure and technique determination during bipedal motion|

---

KR102273591B1|2014-09-29|2021-07-06|현대엠엔소프트 주식회사|Apparatus for wearable terminal and navigation terminal and Method for displaying thereof|

---

CN106767790B|2017-01-12|2019-11-08|厦门大学|The method that human body lower limbs motion model merges estimation pedestrian's mobile tracking with Kalman filtering|

---

US11002820B2|2018-07-30|2021-05-11|7hugs Labs SAS|System for object tracking in physical space with aligned reference frames|

---

CN109949615B|2019-04-08|2021-03-12|中国长江电力股份有限公司|Ship low-speed early warning detection system and method|

---

US11199410B2|2019-04-30|2021-12-14|Stmicroelectronics, Inc.|Dead reckoning by determining misalignment angle between movement direction and sensor heading direction|

法律状态:
2015-11-23| PLFP| Fee payment|Year of fee payment: 3 |

---

2016-11-28| PLFP| Fee payment|Year of fee payment: 4 |

---

2017-11-27| PLFP| Fee payment|Year of fee payment: 5 |

---

2019-11-28| PLFP| Fee payment|Year of fee payment: 7 |

---

2020-11-25| PLFP| Fee payment|Year of fee payment: 8 |

优先权:

---

申请号 | 申请日 | 专利标题

---

FR1362847A|FR3015072B1|2013-12-18|2013-12-18|METHOD FOR DETERMINING THE ORIENTATION OF A MOBILE TERMINAL-RELATED SENSOR MARK WITH SENSOR ASSEMBLY PROVIDED BY A USER AND COMPRISING AT LEAST ONE MOTION-MOVING MOTION SENSOR|FR1362847A| FR3015072B1|2013-12-18|2013-12-18|METHOD FOR DETERMINING THE ORIENTATION OF A MOBILE TERMINAL-RELATED SENSOR MARK WITH SENSOR ASSEMBLY PROVIDED BY A USER AND COMPRISING AT LEAST ONE MOTION-MOVING MOTION SENSOR|

---

US15/105,757| US11002547B2|2013-12-18|2014-12-15|Method for determining the orientation of a sensor frame of reference tied to a mobile terminal carried or worn by a user|

---

PCT/EP2014/077838| WO2015091402A1|2013-12-18|2014-12-15|Method of determining the orientation of a sensor reference frame tied to a mobile terminal furnished with a sensor assembly, carried or worn by a user and comprising at least one motion tied motion sensor|