• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:

    公开号:AT510359A1
    申请号:T0149410
    申请日:2010-09-08
    公开日:2012-03-15
    发明作者:
    申请人:Akg Acoustics Gmbh；
    IPC主号:
    专利说明:

    ···· »II» ·· • * * * I »* * i ·
    t * t * I I • ♦ * * * «• · · · P43780 / rm # * * # * * ·
    • * * · I • · * · ··
    Method for acoustic signal tracking
    The present invention relates to a method for determining the starting position of an acoustic source signal with respect to an electroacoustic transducer arrangement, with a reference database, which has stored reference signals with associated reference positions in the form of angle data and optionally distances of the reference signal from a measurement process, wherein the position determination by comparison of the source signal with the stored reference signals.
    Such a method is known from WO 2009/062211 A1 and enables the position determination of source signals by an arrangement of pressure gradient converters. This method is based on the use of a reference database, which contains the reference signals stored in a measurement process and the associated positions in the form of angle data and optionally distances, the position determination being determined by comparison of the source signal with the stored reference signals. The disadvantage here is that none general statement about the quality of the signals or the detection quality is taken. This means that, for example, in a speech pause, in which no source signal occurs, all background noise and noise occurring act as a source signal and thus the sought source position is detected incorrectly. 20
    The present invention sets out to provide a method of the type mentioned, which is suitable for any type of acoustic signal with noise and transmission pauses of the source signal.
    This object is achieved according to the invention in that, in a first step, characteristic vectors are obtained from the 25 reference signals reference feature vectors and during the position determination during operation of the arrangement from the source signals, the comparison of which with the reference feature vectors results in a similarity curve including both the position and the quality represents the source signal. The advantage of the present invention is that the feature vectors are assembled by the reference feature vectors into a similarity curve representing the position and quality of the source signal. .-2
    Further features and advantages of the invention will become apparent from the subclaims and the following description, which refers to the accompanying drawings. Showing:
    Eig. 1 is the basic flowchart according to the present invention,
    2 shows the schematic flow diagram of the method,
    3 is the detailed flowchart for creating the reference database,
    4a shows the course of the components (vectors) of a reference feature vector and
    4b shows the similarity curve resulting from the reference vector of FIG. 4a based on two different calculation methods, FIG.
    Fig. 5a shows the course of a reference feature vector and
    FIG. 5b shows the similarity curves associated with the noise from the reference feature vector of FIG. 5a and additionally provided with additional noise.
    The individual sequence steps or blocks in the figures are numbered consecutively or uniformly in the respective figures. The essential abbreviations used mean: T reference signal, s source signal, RY reference vectors, Y feature vectors, RD reference database, Q number of positions, q-th source position, M number of microphones, * < - > Transducer input signal of channel m, Hm impulse response of channel m (in the frequency domain), hm amplification factor of channel m (in the time domain), η noise.
    The diagram shown in Fig. 1 shows the basic sequence of the present invention. In a first step, this is done by one, at a defined position, ·· »« ** ··
    Speaker emitted reference signal T (block la) detected by an electro-acoustic transducer assembly. From these m transducer inputs x (m> (block 9a), the reference feature vectors RY (block 2a) are calculated and stored in a reference database RD (block 2 in Fig. 2) .This is repeated for all Q positions S (block Ib) from an unknown position, and the feature vectors Y (block 5) from the converter input signals x (m) (block 9b) are calculated. Thus, each individual feature vector Y can be compared with all reference feature vectors RY, resulting in the similarity curve. From these similarity curves, a resulting similarity curve (block 8) is determined, from which the position Θ is determined.
    The flowchart shown in Fig. 2 shows the schematic sequence of the signal processing with intermediate steps. It can be seen that at least two pieces of information are provided at the transducer input: the acoustic signal (a reference signal T (block 1a) and a source signal S (block 1b)). Indication of the reference position (block 2b) to the respective reference feature RY (block 2a) in the reference database RD (block 2) is required. The basic procedure is as follows: At the beginning, the source signal S is divided into signal blocks (block 3) by means of a windowing (eg rectangle, Hanning) x / (m [and then by means of a Fast Fourier Transform (FFT) (block 4) into the Frequency range transforms X} [f] - FFTN {xl [m]} Subsequently, the feature vectors Y (block 5) calculated therefrom are compared with the reference feature vectors RY already contained in the reference database RD, resulting in a similarity curve (block 6) for each signal block A resulting similarity curve is subsequently determined from these similarity curves, and this resulting similarity curve serves to determine the source position Θ (block 8) with the aid of the reference positions respectively associated with the reference feature vectors RY.
    If, besides the source signals S, there are further true position data from a previous training phase, it is possible according to the embodiment illustrated in FIG. 2 to additionally calculate a localization error (block 7) with the aid of the training data (block 2c), whereby the values in the reference database RD be improved, resulting in an even more accurate positioning. »·« ··· * «· · · · · · · · · · * * * * * * • · · * * * * ι · * - q »* * * ·
    The diagram shown in FIG. 3 shows the detailed procedure for creating the reference database RD (block 2). For this purpose, a microphone arrangement is recorded in a different position from Q Θ with the same reference signal T (block la) via a speaker. From the received input signals, the impulse responses H "are determined (block 10) and transformed into the frequency domain by means of a Fast Fourier Transform (FFT) (block 4). The determined reference feature vectors RY (eg) (block 2a) are stored with the reference positions (block 2b) in the reference database RD (block 2).
    Optionally, the steps in blocks (10 and 4) may also be replaced by another method of acoustic system identification, e.g. by measurement with Maximum Length Sequence (MLS) or Time Delay Spectrometry (TDS).
    The abscissa in Fig. 4a, Fig. 4b and Fig. 5a, Fig. 5b is in each case the azimuthal angle. The ordinate in Fig. 4a and Fig. 5a is respectively a "feature value" (mv) normalized sound pressure level, i. Sound pressure components at the (kidney) microphone positions normalized on the basis of the existing sound pressure at the ball microphone. These values in each case form the feature vector Y over all (kidney) microphones combined for the respective direction. On the other hand, the ordinate in FIG. 4b and FIG. 5b is the similarity (d) between the feature vector Y and the reference feature vectors RY (i & q). ,
    The principle of maximum similarity is shown in FIG. 4a shows the course of the similarity of the three vectors of a reference feature vector / F (0g) (continuous, dashed and dotted), and the values of the feature vector Y are at an azimuthal angle of φ5 = 270 ° in the form of markem (m = T , m = 2, m = 3). This flag vector Y is compared with the reference feature vector RY (0q) in the reference database RD. In Fig. 4b, the resulting similarity curve is plotted from the average of all three similarity curves of the three feature vectors Y based on the p-norm (p-distance) throughout and the angular deviation (cosine) in dashed lines.
    In Fig. 5a, the curves of a reference feature vector / Y (0,) (RY1, RY2, RY3) are shown continuously, dash-dotted and dashed lines. Different sources (si, S2) and one
    omnidirectional noise η are at an azimuthal angle of (ps = -90 ° (η), <ps = 0 ° (sj, St + η), (ps = 45 ° (S1 + S2), φ5 = 90 ° (S2) The corresponding similarity curves are shown in Fig. 5b, which result from the differences of the feature vectors V of two different sources (si, S2) and an omnidirectional noise η, which are marked with markers. the similarity curve of the noise η has a constant course and the combinations of several feature vectors y (s) + S2) and / (sj + η) lead to a flattened course.
    The reference signal T indicates that it contains sufficient energy in the entire reference range, which is later taken into account for source localization. An exponential swept (ES) of the function: x (t) = sin
    0), ie a sine tone with an exponentially increasing frequency, is used as the reference signal T, with the boundary conditions: (3)
    Me &quot; '- 1) 11 dt
    t-T
    A = Τ'ω <ln (β> 2 / £ 9,) T and r = (4) (5) (6) (co. .. start-up sequence, a> 2 .. final-end sequence, T ... ,
    The impulse response Hm for all M channels can be obtained by convoluting the sweep response y (t) of the mth channel with the inverse sweep x (t) '. For this purpose, the measured sweep response jy (Q, as well as the sweep x (t) is transformed into the frequency domain, allowing multiplication of the sweep response Y {w) in the frequency domain and the inverse of the sweep y (w) ~ *. • * - &lt; 7 «Υ (ω) = Χ {ω) Η (ώ) = &gt; Η {ω) = Υ {ω) X (ώ) ~ λ. (7)
    After retransforming the frequency response into the time domain, the impulse response is obtained: H (t) = IDFT {H {w)) (8) (IDFT ... Inverse Discrete Fourier Transformation).
    By means of the impulse responses Hm, the behavior of the array for all Q considered positions and all frequencies can be detected independently of the excitation.
    In order to save storage space and to obtain a smooth frequency response, the impulse responses Hm obtained are limited and faded in, for which purpose a windowing (for example, rectangle, Hanning) is carried out.
    This means that with the reference signals T and the layout of the electroacoustic transducer arrangement b / w. whose impulse responses Hm are the reference feature vectors and the feature vectors Y are calculated in the same way with the source signals S. Depending on the arrangement of the gradient transducers, the reference signals T are multiplied by the respective direction-dependent amplification factors of the M channels of the electroacoustic transducer arrangement, where: hm | Q $ = [hö | 0j, -.-, bjw-7 | Oi] Positioning Θ of the speaker for electroacoustic transducer arrangement in the form of angle and distance indication with the reference feature vectors RY (& q) for Q positions in the reference database RD noted. This process is repeated around the electroacoustic transducer array in, for example, 15 ° steps in Q positions (at 15 ° it follows Q = 24) until the electroacoustic transducer array has received reference signals T from all sides, calculates the associated reference feature vectors RY (Of), and Angular and distance data are entered in the reference database RD.
    The direction to the position of the reference signal T is indicated by two angles: the azimuthal angle cp and the elevation angle Θ, the azimuthal angle φ describing the rotation to the reference signal origin Θ in the horizontal plane and the elevation angle Θ in the vertical plane. The angle origin of the azimuthal angle φ depends on the orientation of the transducer arrangement and is therefore fixed and the angle origin of the elevation angle Θ is normal to the ground plane.
    The reference feature vectors ΛΚ (Θ ") are calculated from the ratio of the five magnitude frequency responses of the kidney microphone to the magnitude frequency response of the omni-directional microphone. Thus, for M-1 cardiac microphones, one obtains for each frequency M-1 characteristic, i. independent of the reference signal T, feature values per reference position Θ. The reference database RD consists of the Q reference positions Θ and the associated m-l magnitude sequence ratios. In the same way, unknown source signals S are detected by the electroacoustic transducer arrangement and, if appropriate, the feature vectors Y are calculated. Whereupon the feature vectors Y are compared with the reference feature vectors ΛΚ (Θ,) with a minimum distance function (MD), whereby a similarity curve is created. This process is repeated for all feature vectors Y and from this the resulting similarity curve is created, which can be determined from the average or even the central value.
    Each azimuthal angle tps has M (number of microphones) frequency-dependent gain factor hm | <,> s, which together form a vector. This means that if he * is known, the associated azimuthal angle &lt; ps is also known. However, in practice, a source position Os is unknown and hence h © j, only received converter input signals Xm are known: 25
    (9) (10)
    It should be noted that the reference signal T is basically always existent. As explained in detail previously, an exponential sine sweep is used as the reference signal T, with the result that this signal is deterministic, analytically describable and reproducible. For detection of the source position Θ, the omnidirectional converter input signal Xo is used, since in a noise-free, omnidirectional microphone h = = 1, it follows that T = Xo or S = Xo Under these optimal conditions: (11)
    «* •«
    * S * - X. (12) RY_ = Y =
    By equation (Eq.) 11 and Eq. 12, the reference feature vector RY (® <,) or the feature vector Y can now be calculated for all m channels: Y = (V;, ..., Ym -])] where Y0 is not taken into account, since Y0 contains no information includes. In addition, the reference feature vector RT ^) is independent of the reference signal T due to Eq. 9 and Eq. 10:
    RY m
    th
    Th m | © s Ojös
    (13)
    The feature vectors Y are determined in the same way as the reference feature vectors R7 (0,), it being noted that the reference feature vectors R 1, R are determined from the stored reference signals T, while the feature vectors Y are determined from the source signals S at runtime.
    The incoming reference signals T are processed in blocks and each signal block is transformed into the frequency domain by means of the Fast Fourier Transform (FFT). By appropriate peak determination (e.g., peak picking), a fixed number p of at least one peak in the respective signal block is detected. If such a peak frequency is not included in the reference database RD, simply select the nearest frequency available in the reference database RD.
    One of the transformations related method of signal decoding in the frequency domain is the bandpass filter bank. The classic example of this is the "Real-Time Analyzer", which decomposes a signal with 30-octave band-wide bandpass filters, with comparatively coarse frequency resolution, the time resolution is very good. In contrast, a digital variant of the filter bank is a cascaded pair of mirror-image complementary FIR filters (Finite Impulse Rcsponse-Filter), low pass and high pass.
    Now, for each of the p frequencies, the similarity for all reference angles is determined, giving a number of p similarity curves. These are then averaged to yield the average similarity curve whose maximum is at that reference position: t closest to the source position Θ.
    More specifically, the similarity curve is obtained by comparing the feature vector / with the reference feature vectors RYi & J in the reference database RD containing Q positions, the p-norm (for p-2 corresponds to the Euclidean distance function) or the angular deviation in same way can be used.
    The p-norm-based similarity simp reads:
    Sim AY, RY} = 1
    l + \ (Y-RY) \ p The similarity of Sinrico based on the wiggle-deviation (cosine function) is: Y1 RY (14)
    Sirn {r, Ry} = n-wi (15)
    Here, the similarity between Y and RY (0f) is given by: (16) S (0q) = Sim {/, ÄK (0q)}.
    Eq. 16 is a function based on the reference position Θ in the reference database RD and is referred to as a similarity curve, since the minimum distance (MD) corresponds to the maximum similarity. Thus, the source position is 0s: 0, = argmax {5, (0)}. (17)
    The resulting similarity curve is composed of the average or the central value of the individual similarity curves.
    In the case of a plurality of simultaneously occurring source signals S, the maximum value of the resulting similarity curve depends on how different the azimuthal angles of the source signals S are. In an extreme case of 180 ° difference, the similarity curve has a course of omnidirectional noise η.
    If there is clear directional information, then the similarity curve in this direction is correspondingly sharp - i. this direction has a great similarity value, while for the other directions correspondingly small values are present. If we calculate the variance of the similarity values for this particular case, then this is large. However, if the direction is ambiguous, there are almost no differences between the directions in the similarity curve, so that the resulting dispersion is small in this case.
    In order to determine whether the determined detection quality is sufficient for the follow-up of the source signal S, a previously empirically determined reference variance is used for the calculated scattering. In operation, the resulting variance is then determined from the resulting similarity curve. Subsequently, the resulting variance is divided by the previously empirically determined reference variance, from which follows: If the resulting value QR is close to 0, the previously detected positions are used to determine the current position Θ of the source signal S. • If the resulting value is QR &lt; 1 and QR »0, there is a well locatable source and this signal block is used to determine the position Θ of the source signal S.
    The determination of the individual limits can be carried out by a person skilled in the art on the basis of a few tests.
    The current direction of the source signal S is determined not only by the currently detected similarity curve but also by the previously detected similarity curves.
    If the omnidirectional noise η increases, the occurring signal-to-noise ratio (SNR) is also significantly worse.
    As low it has proven to perform a preselection of the input data signal. For example, if you want to locate a spokesperson, you should at least have information about whether or not there is any language.
    To improve the robustness of the direction detection algorithm in noisy environments, the reference curves obtained in ideal, noiseless environments are adjusted by applying an offset. That one uses different, dependent on the signal-to-noise ratio (SNR) reference data sets, which are generated artificially and / or obtained by appropriate measurements. »· · · · · · · · · · · · · · · · · · · · · · · · · · · ·. * «« ·· «· * ftft ft · · · _ ♦ ··
    The process according to the invention can be formulated as follows:
    Method for determining the starting position Θ of an acoustic source signal S with respect to an electroacoustic transducer arrangement, with a reference database RD, which has stored reference signals T with associated Q reference positions 0q 5 in the form of angle data and optionally distances of the reference signal T from a measurement process, wherein the position determination by comparison of the source signal S with the stored reference signals T, wherein feature vectors Y are obtained from the reference signals T reference feature vectors R and from the source signals S, and that the feature vectors Y are represented by the reference feature vectors RY to be a similarity curve representing position Θ and the quality of the source signal S. be assembled.
    Method according to the above paragraph, wherein the source signal S is limited in time by means of a windowing.
    Method according to at least one of the above paragraphs, wherein the source signal S is transformed into the frequency domain by means of a fast Fourier transformation (FFT).
    Method according to at least one of the above paragraphs, wherein the source signal (S) is decomposed by means of a bandpass filter bank in the frequency domain.
    Method according to at least one of the above paragraphs, wherein the similarity curve is determined from the difference of a feature vector Y and at least one reference feature vector J F (0f) coming from the reference database RD20.
    Method according to at least one of the above paragraphs, wherein the similarity between the feature vectors Y and the reference feature vectors RY is determined by a distance measure: S = S m {Y, RY}.
    Method according to at least one of the above paragraphs, wherein the determination of the similarity 25 according to the formulation Sim {Υ, / ΪΥΤ = ----. P 1 + IIOr-Änil,
    Method according to at least one of the above paragraphs, wherein the determination of the similarity ··· »···· •» · · · * • «· * ** &lt; '-π- *
    Υτ · RY according to the formulation Simcos {7, RY} takes place.
    Method according to at least one of the above paragraphs, wherein in the case of several p
    Feature vectors Y, the resulting similarity curve is formed from the average of the respective p similarity curves. 5 method according to at least one of the above paragraphs, wherein in the case of several p
    Feature vectors Y, the resulting similarity curve is formed from the central value of the respective p similarity curves.
    Method according to at least one of the above paragraphs, wherein the feature vector Y, which has the minimum difference for all reference feature vectors / F (0f), is used to determine the starting position Θ of the source signal S.
    Method according to at least one of the above paragraphs, wherein a preselection of the
    Source signal S takes place.
    Method according to at least one of the above paragraphs, wherein the current direction to the starting position Θ of the source signal S is determined by the currently detected direction and the 15 previously detected directions.
    Method according to at least one of the above paragraphs, wherein the reference feature vectors RY (&amp; t) are subjected to signal-to-noise ratio (SNR) dependent reference data sets.
    Method according to at least one of the above paragraphs, wherein the signal-to-noise ratio (SNR) dependent reference data sets are generated artificially and / or by measurements. 20
    权利要求:
    Claims (15)
    [1]
    1. A method for determining the starting position (Θ) of an acoustic source signal (S) with respect to a clektroakustischen transducer arrangement, with a reference database (RD) of a measurement reference signals (T) with associated (Q) reference positions (0q) in the form of Angle data and optionally distances of the reference signal (T) has been stored, the position being determined by comparing the source signal (S) with the stored reference signals (T), characterized in that from the reference signals (T) reference feature vectors (RY) and from the source signals ( S) feature vectors (V) are obtained, and that the feature vectors (V) by the reference feature vectors (RV) to a position (Θ) and the quality of the source signal (S) representing similarity curve are composed.
    [2]
    2. The method according to claim 1, characterized in that the source signal (S) is limited in time by means of a fenestration.
    [3]
    3. The method according to claim 1 or 2, characterized in that the source signal (S) by means of a fast Fourier transform (FFT) is transformed into the frequency domain.
    [4]
    4. The method according to any one of claims 1 to 3, characterized in that the source signal (S) is decomposed by means of a bandpass filter bank in the frequency domain.
    [5]
    5. The method according to any one of claims 1 to 4, characterized in that the similarity curve from the difference of a feature vector (Y) and at least one reference from the reference database (RD) reference feature vector (ΛΚ (Θ,)) is determined.
    [6]
    6. The method according to claim 5, characterized in that the similarity between the feature vectors (V) and the reference feature vectors (RY) is determined by a distance measure: S = S m {Y, RY}.
    [7]
    Method according to claim 5 or 6, characterized in that the determination of the 1 similarity is made in accordance with the formulation Simp {y, / F} 1+ | <Y-Äy) t | p.
    [8]
    8. The method according to claim 5 or 6, characterized in that the determination of the similarity according to the formulation Simcos {Y, RY} = Y1 RY

    he follows.
    [9]
    9. The method according to any one of claims 1 to 8, characterized in that in the case of several (p) feature vectors (/) the resulting similarity curve is formed from the average of the respective (p) similarity curves.
    [10]
    10. The method according to any one of claims 1 to 8, characterized in that in the case of several (p) feature vectors (Y) the resulting similarity curve is formed from the central value of the respective (p) similarity curves.
    [11]
    11. The method according to any one of claims 1 to 8, characterized in that that feature vector (V), which for all reference feature vectors (/ Κ (Θ ,,)) has the minimum difference, for determining the starting position (Θ) of the source signal ( S) is used.
    [12]
    12. The method according to any one of claims 1 to 11, characterized in that a preselection of the source signal (S) takes place.
    [13]
    13. The method according to any one of claims 1 to 12, characterized in that the current direction to the starting position (Θ) of the source signal (S) by the currently detected direction and the previously delektierten directions is determined.
    [14]
    14. The method according to any one of claims 1 to 13, characterized in that the reference feature vectors (ÄKf ©,)) with signal-to-noise ratio (SNR) dependent reference data sets are acted upon.
    [15]
    15. The method according to claim 14, characterized in that the signal-to-noise ratio (SNR) dependent reference data sets are generated artificially and / or by measurements.
    类似技术:
    公开号 | 公开日 | 专利标题
    DE112014003443B4|2016-12-29|microphone calibration
    AT510359B1|2015-05-15|METHOD FOR ACOUSTIC SIGNAL TRACKING
    DE60212528T2|2007-01-18|A method of improving near-voice activity detection in a speaker localization system using beamforming
    DE10313331B4|2005-06-16|Method for determining an incident direction of a signal of an acoustic signal source and apparatus for carrying out the method
    DE102017102134A1|2017-08-03|Globally optimized postfiltering with the least squares method for speech enhancement
    DE102016003133A1|2017-09-21|Method for automatically determining an individual function of a DPOAE level map of a human or animal hearing
    DE112016006218B4|2022-02-10|Sound Signal Enhancement Device
    CN106057210A|2016-10-26|Quick speech blind source separation method based on frequency point selection under binaural distance
    DE60108237T2|2006-03-02|RECEIVING SYSTEM FOR A MULTIPLE RECEIVER ANTENNA
    DE60123589T2|2007-08-16|METHOD FOR ESTIMATING A CORRELATION MATRIX OF INTERFERING SIGNALS RECEIVED BY A SENSOR NETWORK
    DE102018109247A1|2018-12-13|Method for accurately calculating the direction of arrival of sound on a microphone array
    DE102013203596A1|2013-09-05|Microphone and method for modeling microphone characteristics
    DE102012025016B3|2014-05-08|Method for determining at least two individual signals from at least two output signals
    DE112017007800T5|2020-06-25|Noise elimination device and noise elimination method
    DE102016225204A1|2018-06-21|Method for operating a hearing aid
    WO1999030446A1|1999-06-17|Method and measuring device to measure the characteristics of radio channels
    DE112018003068T5|2020-03-05|ADVANCED PACKAGE-BASED SAMPLE AUDIO COVERAGE
    DE112010005209T5|2012-11-15|METHOD AND SYSTEM FOR DETECTING OFDM FRAME LIMITS IN A VEHICLE CHANNEL CHANNEL
    EP3275212B1|2019-06-26|Method for analysing and decomposing stereo audio signals
    Lima et al.2016|Low complexity blind separation technique to solve the permutation ambiguity of convolutive speech mixtures
    DE112017007051T5|2019-10-31|Signal processing device
    DE102019217189A1|2020-05-07|Wave field processing method
    EP1176428B1|2007-08-01|Method for improving the signal-to-noise ratio of a sonar array
    EP0787416B1|1998-03-18|Process for measuring the fidelity of stereophonic audio signals and process for recognising stereophonic audio signals coded together
    WO2016030545A2|2016-03-03|Comparison or optimization of signals using the covariance of algebraic invariants
    同族专利:
    公开号 | 公开日
    EP2429214A3|2015-02-25|
    EP2429214A2|2012-03-14|
    AT510359B1|2015-05-15|
    CN102438191B|2015-09-02|
    CN102438191A|2012-05-02|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US5850455A|1996-06-18|1998-12-15|Extreme Audio Reality, Inc.|Discrete dynamic positioning of audio signals in a 360° environment|
    WO2003044516A1|2001-10-31|2003-05-30|Sintef Energiforskning As|Device and method for acoustic detection and localization of defects|
    WO2003101150A1|2002-05-27|2003-12-04|Sonicemotion Ag|Method and device for generating data about the mutual position of at least three acoustic transducers|
    JP4177452B2|2006-11-09|2008-11-05|松下電器産業株式会社|Sound source position detector|
    CN101816191B|2007-09-26|2014-09-17|弗劳恩霍夫应用研究促进协会|Apparatus and method for extracting an ambient signal|
    WO2009062211A1|2007-11-13|2009-05-22|Akg Acoustics Gmbh|Position determination of sound sources|
    JP5397131B2|2009-09-29|2014-01-22|沖電気工業株式会社|Sound source direction estimating apparatus and program|CN103488813A|2013-08-14|2014-01-01|西北工业大学|Optimized design method for layout of multi-component structure system based on P norm|
    CN104732531B|2015-03-11|2017-07-07|中国空间技术研究院|A kind of high-resolution remote sensing image signal to noise ratio curve self-adapting acquisition methods|
    CN105426640B|2015-12-28|2018-11-02|西北工业大学|Method for layout optimal design of multi-assembly structure system based on penalty function|
    CN109238611B|2018-09-27|2020-01-03|广州供电局有限公司|Vibration source detection method and device, computer equipment and storage medium|
    CN109409232B|2018-09-27|2020-04-10|广州供电局有限公司|Noise source detection method, noise source detection device, computer equipment and storage medium|
    法律状态:
    2020-08-15| MM01| Lapse because of not paying annual fees|Effective date: 20190908 |
    优先权:
    申请号 | 申请日 | 专利标题
    ATA1494/2010A|AT510359B1|2010-09-08|2010-09-08|METHOD FOR ACOUSTIC SIGNAL TRACKING|ATA1494/2010A| AT510359B1|2010-09-08|2010-09-08|METHOD FOR ACOUSTIC SIGNAL TRACKING|
    EP11450113.3A| EP2429214A3|2010-09-08|2011-09-06|Method for acoustic signal tracking|
    CN201110264422.6A| CN102438191B|2010-09-08|2011-09-08|For the method that acoustic signal is followed the tracks of|
    [返回顶部]