• 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:
    filter and method for informed spatial filtering using multiple estimates of the direction o and instant arrival a filter (100) to generate an audio output signal, comprising a plurality of samples of the audio output signal based on two or more input signals from the microphone, is provided. the audio output signal and the two or more microphone input signals are represented in a time-frequency domain, characterized in that each of the plurality of samples of the audio output signal is assigned to a time-frequency position ( (k, n)) from a plurality of time-frequency positions ((k, n)).
    公开号:BR112015014380A2
    申请号:R112015014380
    申请日:2013-11-25
    公开日:2020-01-28
    发明作者:Sebastian Braun；Emanuel Habets；Maja Taseska；Oliver Thiergart
    申请人:Fraunhofer Ges Forschung；
    IPC主号:
    专利说明:

    FILTER AND METHOD FOR IHFORMED SPACE FILTERING USING MULTIPLE ESTIMATES OF IRSTAMTÂHEA ARRIVAL DIRECTION
    DESCRIPTION [QOÜ1J The present invention relates to the processing of audio signals and, in particular, to a filter and method for informed spatial filtering using multiple astimtives of the instantaneous arrival direction.
    [0002] The extraction of sound sources in noisy and reverberating conditions is generally found in modern communication systems. Over the past four decades, a wide variety of spatial filtering techniques has been proposed to address this task. The existing spatial filters are optimal when the observed signals are in agreement with the signal model s when the information needed to calculate the filters is accurate. In practice, however, the signal model is generally vicladé and estimating the necessary information is a major challenge.
    roOGã] The existing spatial filters can be broadly classified into linear spatial filters (see, p.eXx, [1, 2, 3, 41} and parameterized spatial filters (see, p.eXx, [S, 6, 7 f 8]) ), in general, linear spatial filters require an estimate, of one or more propagation vectors or second-order statistics (80S | seuòad-order statistics; gives one or more desired sources plus 80S of interference. Some spatial filters are designed to extract a single source signal, both naked and reverberated, (see,
    eg, [9, 10, 11, 12, 13, 14, 13, 10]), while others were designed to extract x sum of the two most significant origins (life, eg, [ 17, ΐ3]) * The previously mentioned methods require cement · anterior of the direction of an or. more desired sources or a period in which only the desired sources are active, either separately or only 1>
    (00δ4 a disadvantage of these methods is the inability to adapt quickly and sufficiently to new situations f for example, concurrent mobile devices or loudspeakers that become active when the desired source is active. Parametric spatial filters usually have. based on a relatively simple signal model, for example, the signal received in the time domain ~ fragrance consists of a single plane wave plus the diffuse sound, and is calculated * based on estimates simulated by the model parameters. parametric parameters are a highly flexible directional response, a comparatively strong suppression of diffused sound and interiors, and the ability to quickly adapt to new situations. However, as shown in [19], the underlying model of single plane wave signal can be easily violated in practice, which strongly degrades the performance of parametric spatial filters. that the state of the art parametric spatial filters u.ti read all available microphone signals to estimate the model parameters, while only one



    / fcnt ^ ís) sound (s), and being adapted to set the weighting information to nothing - one of the plurality of cover positions · frequency, depending on the arrival direction information of am or more component te) of the sound. of one or more sound source (s) from said temp-frequency position or depending on the position information of one or more sound source (s) from said time-frequency position <
    Ϊ 0009] In addition, the filter complies with a ge rsdardo signal d® salda to generate the audio output signal, generating, for each of the plurality of tpu-sequence positions, bare of the worst output of the audio output signal samples. , which is attributed to said time-frequency position, depending on the weighting information of said time-frequency position and depending on an audio input sample, being attributed to said position of the time required in 1 st, oa of a finger .1 s or more 1 in the microphone's front end.
    ÍOOÍOj The applications provide a filter «spac.is.1 to obtain a desired response for, at the most, simultaneous active sound sources f-, .0 the provided spatial filter is obtained by reducing the diffuse-more-noise power in the filter outlet subjected to linear restrictions .1, As opposed to the concepts of the state of tactics, linear .1 are based on the estimates of the · narrow band arrival direction xnstant àn ® a. A .1 in d 1 s s, in v a s a t a d e a d a d i c a of diffuse-to-noise / power, diffuse are provided, exhibiting a time-to-spectral resolution that is sufficiently high to achieve both de-reverberation and noise reduction.
    [00111 According to some applications, these concepts are provided to obtain an arbitrary spatial response desired for Λ at most, L fontessurring simultaneously active by time-frequency instast <for this purpose, the instant parametric information (IPX i. instantaneous parametric .information) on the acoustic scene is incorporated into the design of a spatial filter resulting in an informative spatial filter ''.
    [0012] In some applications, such informed spatial filter, for example, combines all available microphone signals, based on complex weightings, to provide an improved output signal »(001.3] Applications sound accord> the spatial filter For example, it can be performed as a special filter ...... of minimally restricted variance (LCMV) linearly oonstrained minimum variance) or as a fi 1.1 rod í en er mu 11. íaa π at 1 pair am is tr 1 c o.
    Γ0014] In some applications, the provided spatial filter provided and, for example, obtained by reducing the more diffuse self-noise power subjected to linear constraints L <
    (GDIS] In some applications, unlike the prior art, the L-linears are based on estimates of the incoming direction of the (DOA) direction ti ti-of-arr1va1J instantaneous, and the resulting responses to the 1 DOAs correspond to the specific desired directivity. .
    [QOl 6} Hemdi® so, nine dockers for the s. '. a 11 at i a y a y de s ad ill ê ru ide f pore xs mp 1 ο, the index of diffusion-to-rnldo (DNR · dif.fuse-fo-noiae ratio), Síàó forneeloos azibando a shallow temporal lotion and high spectral suficientsments, for example, to reduce both reverberation and noise>
    [0017] In addition, a method for generating an audio output signal, comprising a plurality of samples of the audio output, the bass am doia or more microphone input signals, is provided, the audio output signal audio and the two or more microphone input signals are represented in a time-frequency domain, characterized by each of the plurality of samples of the audio output signal being assigned to a temperature-frequency position of a plurality of frequency-frequency, the method comprises:
    ~ Reoe.bimento, for each of the plurality of fit time-sequence positions, of the arrival direction information of one or more components (s) of the sound from one or more sound source (s) or position information one or more sound sources (s),
    - Generation of weighting information for each of the plurality of tempc-frequency positions, depending on the arrival direction information of one or more components (si of the sound of one or more sound source (s) of said position. time-, frequency or depending on the position information of one or more sound source (s) of the


    8/52


    tempc-frequency position ..... fk, n / ... depending on the weighting information of said ..... time position · weakness ...... f.è ,, ..... .n.1 and depending on u you give audio input samples of no two or more microphone input signals # namely # for example # depending on one of the audio input samples of no two or more microphone input terminals # which is assigned to that time position * frequency fk z nj «[0035] For each master of the audio output signal, w ow the pa rt 11 to the location of the time. (1., n; ... the weighting generator1010 recently generates the individual weighting information, the leO output signal generator generates # then # the sample of the audio output signal from the freetime time position considered easy # n) based on the weighting information generated for that time position ™ frequency x In other words # a 'new weighting information a calculated by the weighting generator 110 for each time-frequency position # which bears, sample of the audio output signal must be generated, [G03.6] Ac generating weighting information # weighting generator 110 is adapted to consider information about one or more sound source (s).
    [0037] Fo «& xempl o # o g e r a of the po nde r a goes 110 can consider a position of a source of sorption, They are an application # the weighting generator can # still # consider a position of a second sound source.
    [0033] Gu # for example # the first sound source can emit a first sound wave with a first sound component. The first sound wave with- the first sound component · arrives in a microphone and the weighting generator 110 can consider the direction of arrival of the first sound component / of the sound wave »In this respect # · the weighting generator .110 considers the information in the first sound source »In addition # the second sound source can also use a second sound wave with a second sound component ·» The second wave is odd with the second common sound and reaches the microphone and the sound generator. ponderations. 1.10 can consider the direction of arrival of the second sound component / second sound wave. In relation to this - the generator of PCnderagces 110 ccnsidcra # still - the information about the second sound source. [003 $] Fig. I.b illustrates one. possible application scenario for a 100 # filter according to an application. A first sound wave with a first sound component is in tide for a p r 1 m and i roelto-falentelil. (·. · Mapri me iraf on te sonora) and arrives on a first microphone 1.11 <Direction of arrival of the first sound component (* = direction of arrival of the first sound wave; # on the first microphone # 111 is considered. In addition # a second sound wave with a second sound component is emitted by a second loudspeaker * speaker 123 (a second sound source) and arrives at the first microphone 111. The weighting generator 110 is capable of considering # yet # the direction of the arrival of the second sound component in the first microphone 11.1 to determine the weighting information, in addition # the direction of arrival of the sound components (direction of arrival of the sound waves) in other microphones, such as minophone 112 can also be considered by the weighting generator stop taking to 1 nfo rmation ροή de ra ç ã ο. » [0040] It should be noted that the sound sources may, for example, be physical sound sources that exist, physically in an environment, for example, speakers, musical instruments or a person speaking ·.
    [Ó ó 41] Auto enter: de ve se robs herb than the sources of the mirror image are still sound sources. For example, a sound wave emitted by a speaker 122 can be reflected by a wall 122 and the sound wave then appears to be emitted from a position 123 being different from. position of the speaker that, in fact, emitted the sound wave. Such source 4a mirror image 123 is, still, considered as a sound sourcex The weighting generator 110 can be adapted to generate the : weighting information, depending on the direction of arrival information referring to a source of the mirror i.macem or . depending on position information in one, two or more sources of the spread image.
    [0042] Fig. 2 illustrates a filter 100, according to one. application and a plurality of microphones 111, 112, 1'13 ,,. ,, lln, 14a application of Figure 2, and the filter 100 further comprises nm. filter bank 101. In addition, in the application of Fig. 2> the weighting generator 110 òrmvgresnde a module of the calculation camputeclona1 of the information 102, the module of the calculation oomputacian.al of weightings 10 3 and a module of the transfer function selection . 104.
    ccndueidax More specifically, the weightings w (A, n] are, for example, calculated in the case of the information for you; you try instantaneously (I PI] J (xçn) $ C based on a desired transfer function corresponding to G ( x, a).
    [G047J 0 .onio computational module nil information is configured to calculate 10'2 XFX from the microphone signals x (k, * AI to FI} describes the characteristics especificasdo sinalecs camponantes noise included in the microphone signals {χ 1> n) pars ο given time-frequency instant tip η)>
    (0048j Fig. 3 illustrates, a weighting generator
    110, according to an application. Q weight generator
    110 comprises a computational information calculation module 102, a computational weighting calculation module 103 and a transfer function selection module 104, [004 9] As shown in the example in Fig. 3, the IPX mainly comprises the instantaneous arrival direction (DOA) of one or more components (s of directional sound (for example, plane waves), for example, calculated by a DOA 201 estimation module, [0050] As explained below, the information of
    DOA can be represented as -an angle (for example, by (azimuth angle çif, o.), By elevation angle n) by a spatial frequency (for example, by μ [A i çfà, n)]; , by a phase change (for example, by a [x í
    n) J} for a delay time. the time between microphones, for a vector

    For example, be the component: of stationary noise / slowly varying noise components and the second noise information may be information about the components of noise and slowly varying «(0059] In one application # the weighting generator
    110 is can be used to generate the first noise information (for example, the information about the noise components is not included in the ids / slowly 1 or variants Γ, using, for example, the predefined statistical information (for example, For example, information about a spatial coherence between two or more input signals from the miorophone resulting from non-static noise breaks), and characterized by the weighting generator 110 being configured to generate the second noise information (for example, infor- mation on the components of noise that are leutely variant / stationary) without using statistical information.
    (0060] Regarding noise components that change quickly, the microphone input signals alone do not provide enough information to determine the information about those noise components. Statistical information is, for example, additionally necessary to determine the information regarding to fast-changing noise components <
    [0061] trailer, referring to the noise components that do not change or change slowly, statistical information is not necessary. to determine information about these noise components. Instead, it is sufficient to evaluate the signs of the myorophone.
    [0052] It should be noted that the statistical information can be calculated, excluding the estimated DOA information, cifiform shown in Fig. 3 <. It should also be noted that IPX can also be provided enter nascent a, For example, a DOA of s <»(a position of sound sources, respectivelyj can be determined by a v1ded camera along with a r algorithm facial echo, assuming that human speakers form the rescue scene, [0033] A transfer function selection module 104 is configured to provide a transfer function dfb, n;. complex potencíalments] Gfk, aj of Figure 2 and Figure 3 of the key to the desired response of the system that received the IPX (for example, current parametric) dhsp g or example, G {'à, M can describe an arbitrary withdrawal pattern of a desired space microphone for improving the signal in mono reproduction, a DOA-dependent speaker gain
    for reproducing the a 11 o - f a 1 an t e, o u a function of related transfer to the head (HsTF} for reproduction binaura1. (0064] Dave be c b s e r v ate qu e, q e r a 1 m e n t e, a s
    The statistics of a recorded cane sonet vary rapidly over time and frequency. Consequently, the IPU JUtn) to ajJ optimal weightings corresponding to w {.k, n) are valid only for a specific top-sequence index.
    19/52 and, the information is recalculated for each enk. Therefore, the systana can adapt to r - s and instants after the current recording.
    (00651 It should also be noted that the M input norephants can either make a single natris on the niorophone or be distributed to make various natrires in different locations. In addition, the IPX Jtkon) can understand the position information, rather than the information of DOA, eg employing the positions of sound sources in a three-dimensional environment * Because of this, spatial filters can be defined not only by filtering spherical directions, as desired, in the three-dimensional spatial regions of the recording cane.
    (00661 All explanations provided with respect to DOs are equally applicable when an inferred position of a sound source is available. For example, the
    i.nfraction of position can be represented by a DOA (an angle) and a distance. When such a representation of the position is used, the DOA can be obtained immediately, from the position information. Or, the position information can, for exenpic, be described by the agreed x, y, x. Then, the DOA can be calculated on the basis of the position information of the sound source and based on a position of the ncróforie, which records the respective signal of the input ni or o.fone.
    [vúál] Other applications are described below.
    fOD68] Some applications allow the recording of sound ospaoi a selective magnet with reverberation and noise reduction. In this context, applications for the application of spatial filtering for signal freehand in terms of source extraction, de-reverberation and noise writing are provided. The purpose of these applications is to calculate a current Y (k, rq that corresponds to the output of a directional microphone with an arbitrary withdrawal pattern. This means that co dimeions 1 {for ex mp 1 o, a single on d. A pi ana) is it up to me presérvádò, as desired, depending on your WFD while the diffuse sound or the noise of the microphone is s up laughing. Depending on the application, the spatial filter provided combines the benefits of state-of-the-art spatial filters, inter alia, providing a high directivity index (DI) directivity .index) and situations with a high DNR, and a high noise gain white (WG i unite noise gala.), otherwise. According to algesian applications, the spatial filter can only be restricted linearities, allowing a quick corputation calculation of the weights. For example, the transfer function n) of Fig. 2 a of Fiç, 3 can, for example, represent a withdrawal pattern. directional microphone.
    [OOnõl below, a formula of the problem is provided <Bntao, the applications of the computational weighting module 103 in the computational calculation module of the IPX 102 for the recording of spatially selective sound with reverberation and noise reduction are provided. In addition said, the applications of a selection module
    Corresponding TF 104 are described <
    í 0 0 7 0] Fm primé 1 xo 1 úga xá f ô.rmu la of the problem is closed <of the matrix of the omnidirectional miorofon.es Zf 1 oo to 1i ized am Φ, f , " P araca da (i f ft) ρ xes uw ~ if a field of sun's is composed of flat waves D <M {oom dl rsei ona 1) that propagate in a field of isúxdpico and espauialmeóté bomegênso sound fuse, The signals of the micxófçne x (i, n) can be written as · x (Í t 0 -] jP xpi <a.) è xa (L ft 1 4 'XgÇís η).
    <:: s 1 í [0071 j where xH.k z n) · «r. <dü ♦. x <
    dU p cempresris the microphone signals that are proportional to the preset of the i-th flat wave, xü ê the measured non-stationary noise (for example # diffuse sound) and χ Λ Τ ; rd is stationary noise / slow noise on the variance (for example, au torruí do do mi and rophone), [0072} Assuming that the three components in the
    Formula (2) are mutually unrelated, the matrix of the spectral power density (PSD | pomer spectral density; of the microphone signals can be described by
    4> (À s) ~~ E {x. (À · η) χ Η (ί .0 ')} £
    t W> d (E n ~ cm (A 0 Γ, ηΗ [0074] Here # ^. (2, n) is the PSD matrix of stationary noise Z. slow & variant noise and ó <s (t, n ) is the expected power of the non-stationary noise, which can vary rapidly from time to time. The element 1 / -th of the cosrence matrix «X / pi], denoted by yiq (^) r coherence between the ieg microphones resulting from the non-stationary noise For example, for a diffuse champion sphericamante iaotropioo, y'm (ê) sincLK x.í 4 ) [20], with wave number κ o xm m H <, 7 - dj H 0 element ig-th da. Matrix of coherence F R [.â) is the coherence between microphones iej resulting from stationary noise / slow variant noise. For the microphone's performance, n) ~ p rs [ê, ai X, where I is an identity matrix and% (ê, n) is the expected power of self-progress.
    [0075] The directional sound x> [i, ή) in (2) asks to be written as
    X [(àt n) ™ a [fe | pg (kn) | Xi (Kn <di),, 5 , [007S] where n) is the acimute of the DOA of the 1-tn plane wave (p- ~ 0 denoting the wide side of the matrix) and e [ii çu (Àz n)) [ s-ι ík | a)] ... eeíà 'I ¢ 1 (^ · Ρ) ίΓ' ê α propagation vector .. The element i * th of a [& i pjÇxt nj 1, ipiê i çquc n) j - exp {, rr * sinn)}, [0077] describes the change in the fa.se of where 1-th plane from the first microphone to the .i-th microphone »It should be noted that xo ;; í Hdl - dl I i is equal to the distance between the first microphone and microphone f ~ th.
    [0070] the angle dad · * I pi. n)] «AHÍ ΐ Ç.HÀ Z η)] is generically referred to as spatial frequency. The DOA of the J-th wave can be represented by g.ü, η), aHd ia)],

    25/52

    d ; : The% u .n, dj <Xs. (Kz dj.)] Comprises the propurential signals to the sound pressures of the flat waves X in the reference microphone, Note that 4mà ú) is a matrix dia go na 1 f es q ç e os e 1 eme ή t d gum isdiag {Φ s (k, n J} ™ íõiíà., rp. v ç £ . (k, η) P are the potãnoxas of the flat arrival steps> In order to have u cent roll over the signal distortions introduced, we can include the diagonal matrix Λ {1., n) comprising time and frequency dependent control parameters diag {Á} ~ Àd, n] A> ♦ »· Ρ (1 , η) P, that is,
    Opwpl A »». [g À ^ (lgn) w | H A (Ç ») # S (A«) A M (k, n) w] is w H Φ »(Ρ H w · ό '/ ο : // d ··· /// j / Linoogx / fo] / [0101] n solution to the suzimisation protdáka in (1 ) f given CsAkz A, à w »[A H A (k <z / Í # s (PhjA 4- ^ J ^ jAAÍAx u ^ sÇk ^ pg .
    [21)
    Í0103} this is identical aw = x Φ ' 1 AjA' 1 ^ -a Α Η Φ ” ι Α]“ ^ í22 .
    (0104] It should be noted that for Α ' Λ ~ 0, the LCMV solution in (14) is obtained. »For A ® X, the Wiener multi range filter is obtained» For other values Λι ,,.! · .. (k, xü, the amount of distortion of the corresponding source signal and the amount of residual noise suppression can be controlled, respectively! Therefore, define '{, kn}, from the foot of the information in detail. for the tricadisp on 1 ve 1,
    weightings can prune. be ca Ion la da *. The DOA of the calculated A plane waves; in siddula 201, they can be obtained with known narrow band EGA estimators bee EEFRIT [221 or MUSIC i23J raise or other eatleaders * of the prior art. These algorithms ask for a # pot example, the animate angle g (k, n}, the spatial frequency p (ki ç (.k z a)}, the phase shift a [kjn} 1 or the veter of p re pa g now a [k Γ p (k, n) ί for era or ma 1 only on wed ch in the matrix A. DOA estimate will not be discussed further, parents the DOA estimate itself is well known in the art.
    [01x1] In the following, the estimation of the noise-diffuse indication (DMR) is described. In particular, the DES estimate of input 1 * [2. n}, that is, a reaction of module 202 in Fig O, is discussed The DMA estimate explores the DOA information obtained from module 201. To estimate k (k, n), an additional spatial filter, which cancels out plane waves 2, so that only diffuse sound. whether it is eapte.rado, it can be used.The weights of this spatial filter are adjusted, for example, by a wg nt a of the matrix> · that is, xvg ™ arapuin w ^ w W ft X v:: - CS.: .-: 4 :: .- 0. ..................... --- 4 <........ 4: 4 <4 4: 4 4 : 44 .-: -: 4.4: ::::: 4: 44 : 44 4% · <· ». ·;>
    [0112] submitted aw H aU-1 <a> (/ g η ή - ü </ c (L 2 .... J,).
    4ji4 [: [: 44: [4-4: 44; çgW h U; A: i pp (k, n); ™ L, 7 iOU3] The constraint {27} guarantees nonzero weightings. The propagation vector at [2 i çç (k, · n)] corresponds to a
    [0118) where w 't is defined in the previous paragraph,
    It should be noted that the PSD matrix matrix stationary / noisy lent variant lover Φ, ρ.Χζ a) can be taken during silence (that is, during: the absence of signal and non-stationary noise), that is, <> «(D, u) - E (xÇn (dw)},,, $ í v [01lã)« that the expectation is approximated by the average on the structures of silence, n. The structures of silence can be detected with the state of the art methods »[0120] Next, the estimation of the P3D matrix of the unwanted signal (see module 203) is discussed.
    [0121] reads more than 2 0.0 of s i n a 1 1 n d e s e j a d (r u i d u stationary / slowly variant plus non-stationary raids) $ 42, n) cannot be obtained nor
    Φ υ (Αηη) «Á, ^) (Ψ (Αη«) Γ <ί (Α :) 4 Γ η (Α ')) <
    {33) [0122] or, more generally, with #u (Ai d) == ÇdPi n) ia (£) 4 · n). $ i ;. ';/;>'·71:.' . · : ' 1 · ; 1: s ·· : ::. · K < ·· 7 z · -::.; Λ << : · : · [0123] where Γ * (ί) ar í; (á) are available as a priori information (see above). DdR $ {2, n) f the power of stationary / slow lover noise. ri ante ò «(xt n) and other necessary quantities may be calculated, as explained above. Thus, the estimate of n) explores the DOA information obtained by module 201 <
    [0124] Next, the estimate of the PSD matrix of s i n a1 (see j 2 0 4 modulus) is s cr i t a.
    [0120] The power dm n) of the plane waves of

    then, obtained by - arpnún w H Φ η (U ) www «w {301 [013.3] Bsistá loves solution da. closed form for (38) (1) which allows a quick computational calculation of. »&. It should be noted that this filter does not necessarily provide the broadest 151.
    [0136] The second spatial filter is known as the robust superdirectly divided beam former (SD 1 super <if recti ve) at reductions. power of the diffuse sound in the filter output [that is, it increases the Dl] with a lower link in the WIG [24] <The lower link in the WNG increases the robustness to the overpricing errors and limits the amplification of the self-noise [24] <The optimal vector of weighting that reduces the MSB between (7) and (3) subject to (M) s satisfies the lower bond in the WG is then obtained by w d ™ arp mm w H Φρ (Α a) w W .. .......... · ...... ;: >
    w K r.-en / x) w (x [01371 is subject to a quadratics constraint w 5 * w <
    The parameter, 8 “'defines the OG .mloime and determines the reachable Dl of the filter. In practice, it is generally difficult to find an optimal disadvantage among enough in
    s.it short SN.R, and a Dl sufi hi fearfully high in. high SNã situations. In addition, solving (3.9} leads to a non-convex optimization problem due to the quadratic constraint, which is time consuming to solve. This is problematic space, since the weighting vector
    37/52
    robust ND falxes (we. A Fig. 8 .. (. b) mux fra the ifNGs cox answered. Du r anta aa i1 ê n ex o, the proposed spatial filter (jadswè trade line reaches w high MG, whereas, during signal activity, WG is relatively low.
    [0144j Fig. 8: the Dl and c WG donates filtered compared spaces. For «g, the minimum WNG was set to -12 dB to make the spatial filter robust against the [Olãõj Overall, aFig. 8 shows that the special filter proposed by you combines the advantages of both existing spatial filters: during the silence parts, a maximum WNG to the furnscldc, leading to a minimum amplification of outer noise, that is, high robustness.
    [0148] During signal activity and high reverberation, the self-noise is usually masked, a high Dl and ferneoido (at the cost of a low WNG) leading to an excellent reduction in the diffuse sound. In this case, even smaller NNGs are to be saved.
    [0147] Note that, for the higher frequencies (f '> 5 kHs), all spatial filters perform almost identically, since the coherence material Χύ (Ι') in (3âj is (12) and approximately equal to a matrix of identity.
    [0148] Following, the directional restrictions i n s t a n t â n and a. s are and on s i r of s.
    [81481 For this simulation, it is assumed that no prior information! on and be available. The DGAs / * /
    Ρι (2, n) and ÇíU, n) are estimated cos ...... ESPRIT <Thus, the constraints (9} vary over the top. Only for the robust ED beam former loves a single constraint. And time invariants (9 ) f corresponding to the fixed viewing direction of (A ::: 33% is used- This beam former serves as a reference.
    [0150] Fig. $ Describes the estimated DOA ζ <· (Χ ' ζ n) and the resulting gains G [.fe I çg (á, n) j. In particular, Fig. 9 illustrates the estimated DOA pit, a) and the resulting gain iG (kim ·. (K, n) 1%. The arrival plans wave is not attenuated, if the POA is within the sword window in the Fig. 4 (solid line) On the contrary, the power of the wave is attenuated by 21 dB.
    [0151] Table 1 shows a performance of all spatial filters (* not required). Values in parentheses refer to time-invariant directional restrictions, values outside parentheses refer to instantaneous directional restrictions. Signals were weighted by A before calculating SIR, SRR and SSKR.
    s nr # t pB [ 111111111 ' PncU s - (0-1 :) I · 28 (29) Α1.Β] 1 (There) d: 21 (32) ... V f ....; Ç 33 ÍSU 2i) (L.T)23 (SB) : 3 (-1; 2: 2 (2i) 2 1 • 2 0} dd 23 [3% : í (4] <li : % rrlr (2b) 2.1 (1.0)
    Table 1 [0152] In particular, Table 1 summarizes the overall performance of spatial filters in terms of signal-to-interference ratio (SIR i Aig-nâi-to-interference ratio), Signal-to-reverting index (SRR signal- all-
    considered and a similar problem is formulated. The stationary / slowly varying noise corresponds to the unwanted microphone self-noise, while the non-stationary noise corresponds to the desired diffuse sound. Diffuse sound is desired in this application, as it is most important to reproduce the original spatial impression of the recording scene.
    [01531 Next, the directional sound reproduction
    Xjil, n, d,;,; without distortions of the corresponding DOA çj (r, n) must be obtained. In addition, diffuse sound must be reproduced with the correct energy from all directions, while the microphone noise is suppressed. Thus, the desired sign ¥ (A, a) in (7, is now expressed as
    11/1/11111 / rl / · '/ 111 ·) / i //////// P / jú /: /// js ///// ^
    U (/ an) a »Xpl.n. di) 'è Gá (âçn) Xrí. (&<ó, d), in iãOj [Olõdj in gu-s n) is the signal of channel 1-th of the asm reproduction system (i »* (1, < .., GH, X 3 < ,. (, 5, n, d) is the diffuse sound measured at an arbitrary point (for example, on the first dj microphone) to be played by the speaker ie ί1 <(χ, n ) is a gain function for diffused sound to ensure correct diffused sound power during playback (gsralmente tò (.k, u) - i / Ç.V; ideally, Xmi signals (A, a) have the power of the correct diffuse sound and are mutually uncorrelated to channels i, ie., », ·., I : i.'í 11. aj, ( í lu n) | ™
    U 'Oé oíúto rr.ndf' ;.
    [01S7j (41)
    The transfer functions G; [À I d> U, n)]
    / V A. (. 4 .3)

    / X- ·
    corresponding or a characteristic or characteristic of a corresponding feature.
    [01t2] Decomposed inventive inventory can be stored in one. digital storage model or can be transmitted by a transmission medium, such as one. naked wireless transmission medium a tired transmission medium, such as the 1st te rn and t * [0170] Depending on certain implementation requirements, the applications of the invention can be implemented in hardware or in software. The implementation can be carried out using a digital storage medium, for example, a floppy disk, a DVD, a CD, a sDM memory, a PÁOpç an EFHOd, aná EEPROM or a FLASH memory, having 1egible level control signals stored in it (or are able to cooperate · with a programmable computer system, so that the respective method is carried out>
    [0171] Some applications, according to the invention, comprise one. non-transient data carrier having electronically readable control signals, which are able to cooperate with a programmable computer system, so that one of the methods described in this document is performed, [0172] In general, the applications of this
    Invention can be implemented as a product of the computer program with a program code, with the program code being operative will perform one of the methods when the product of the computer program is run on a computer »The program code can # by example, be the rma renado in a 1ogive1 transporter per ma guina>
    fG173J Other applications comprise the computer program to carry out one of the methods described in this document, stored on a legitimate conveyor by means of a <
    101741 In a nutshell # an application of the inventive method is therefore # a computer program having a program code to perform one of the methods described in this document, when the computer program is run on a computer.
    [0175] Another application of the inventive methods is, therefore, one. data carrier (or a digitally generated medium, or a computer-readable medium) comprising, recorded, the computer program to perform one of the methods described in this document.
    [0176] Another application of the inventive method is, therefore, a data stream or a sequence of signals that represents, the computer program to perform one; the methods described in this document. The data stream in the signal sequence can, for example # be configured to be transferred via a data communication connection, for example # over the Internet.
    [0177] Other. application comprises a processing means, for example, a. computer or a logical device pro gram ama 1, -configured or adapted to perform one of the methods described in this document.
    [0178] Dutra application comprises a computer, in which the computer program for real irar one of the methods described in the document is installed on it.
    (0179] £ m some applications, a programmable logic device [for example, the arrangement of the door frames orogr amá ve 1 s} can be used 1.1 not compatible with 11 za ra 1 g ama am all the functionality of the methods described in this document In some applications, an arrangement of arogrammable field gates can cooperate with a microprocessor, in order to implement one of the methods described in this document .. In general, the methods adopted are preferably performed by any hardware device.
    [0180] These applications described above are not merely illustrative for the principles of the present invention. It is understood that the modifications and variations of the provisions, and the details described in this document, will be evident to other experts in the art, the intention of the invention, therefore, to be limited only by the scope of the attached claims and not by the specific details presented in the form of description- and explanation of the applications in this document.
    hEFDRêpCXéS [01811 [lj J. Goods st y> u <Chen, and Y <Huang,
    Microphone Array Signal frácsssing. Berlin, Germany: Springer-Verlag, 200B.
    [0182] Γ2] S <. Duel o, S. Gannet, M. Hoonen, and A.
    Sprint, '' Acoustic beamfcrmlng far hearing aid applications, in handtool · on Array Fracassfng and Pansor Mataorks, 8.
    Haykin and K. Ray Liu, Eds. Filey, 2008 # ch. 9.
    [GIB3j [3] S. Gernot and 1. Cohen, Adaptive beamforming and postfiltarlng, in Springer SandhooR of .Speech Proceasing # J. Benesty, M <M. Sondhi, and Y «Huang,
    Eds, Springer-Cerlag # 1008, oh. 47.
    [0184] [4] G. Benesty, J, Chen, and £. A. F.
    Habets, Speech Fa à a n cam /; t In rhe S TFT Dona f a, sen
    SpringerBriefs in Electrical and Computer Engineering.
    SprInge t-Ver1ao, 2011.
    [OlBSj [5] X. lasher, M.. Seltzer, and A <hoero.
    Hletup · One a rray fur baadeer with spatial noise sυppress a r, * in Frac, ifinfeh International dorlshop on Aeoustio, Fcho and Fbisa Contrai fldAxdCJ <Eindhoven, The Netherlands, 2005.
    [0186] [6] H. Kai Unger, G. Del Galdo, F. Kusch, D.
    Manne, and R. Schulta-Amling, Spatialfiltering using directional audio coding parameters, '' 'in Proc. I.S.sF Inti. Conf, an Acoustics # Speech and El Seal Processing flCARUÇ, Apr. 2008, pp. 217-220.
    [018 7] [7] M. Ra 11.1 ng e r, G. D. Ga I do, F. R oe ch, and
    0. Thisrgart, Dereverberation in the spatial audio coding domain, '' 'in Audio. Engineering Society Convention 130, Loudon DK, May 2011.
    [0188] [6] G <From 1 G to 1 do, 0. T h 1 argart, T ♦ be 11 er, and E, AP Habets, Generating vix'tua.l microphone signals using geometrical information gathered by distributed arrays, in Frac. .Eands-Free Speech iCOmríoní cation and
    9/52
    employing aiganvector-based transfer function ratios estimation, · J&SS Trans. Audio, Speech, Bang. Precess., Vol. 11, no. 1., pp. 20 (9-21.9, Jan. 2011.
    [01951 [15] E, A, , Habets and J. Benesty, Joint dereverberation and noise reduction using a two · stags beamfurmiug approach, in Free. Hands-Free Speech Comm F.n 1 ca t ion and Mi or opAsn and Ar rays [PSC1W »2 011, pp. 101 [0190: [16] M. Easeska and E. A. P. Habers, MMSEbased blind source extraction in diffuse noise fields using a complex none rance -based a priori SAB estimator, in Prob. Inti, FforksAop Aeonst. Signal Snhancament [THASFCJ, Sep. 2012, [0197] [17] G. Renven, S, Gannet, and A Cohen, '' Dual source transfer-function generalized sidelobe canceller, ISPs Trans, Speech Audio Process., Vol. 16, no. 4, pp. 71.1-727, May 2000.
    [0109] [18] 8- Markovich, S. Cannot, and X. Cohan,
    Mul.tichamiei eigenspace beamforming in a reverberant noisy environment with multiple interfering speech signals, IE.HE Trans. Audio, Speech, lang. Process .., vol. 17, no. €, pp. 1071-1000, Aug, 2008.
    [0'19.9] [1.9] O. Thiergart and EA, P. Bafoeta,
    Sound field model violations in. parametric spatial sound processing, in Proc, Inti. .Pori shop A coast. Signal Annas semen t ilWAEHC /, Sep. 2012.
    [0280] [20] R. K, Cook, R. V. Waterhouse, R. D,
    Bexendt.3. Edelman, and 9), C. Thompson Jr., '' Heasutementof correlation coefficients ....... in reverberant sound fields, * The
    Journal of are Acoustical focisty of America, vol. 27, no, €, pp. 1072-1077, 1955.
    (0201] [21] 0. L. Frost, HI, As algorithm, for linearly constrained adaptive array processing, Proc. ISSB, vol. 60, no. 8, pp. 920-035, Aug. 1972.
    [0232] [22] A. Roy and T. Kailath, ESPRITOS t ima ti on. of signa 1 pair love t was via rotational 1 overt a nee techniques , i Acoustics, speech and SignalProcessing, IPSE Transactioaa on, vol. 37, no. 7, pp. 714-995, July 1999.
    [0203] [23] B. Rao and K. Hari, Performance analysis of root-music *, in Signals, Systems and Computers, 1963. Twenty-Second Asilcmar ÜOxtTeranee on, vol. 2, 1983, pp. 575-502.
    . [0204] [24] A. Cox, a. M. ieskind, and M. d. Owen,
    Robust adaptive beamforming, lEEn Trans. Aooust., Speech, Sligna.l Process., Vol. 35, nc. IO, pp. 1355-1375, Oct. 1987.
    [0205] [25] J. B »Allen and 0. A. Barkley, Image method for efficiently simulating small-room acoustics, * 'J.
    Acoust. See. .am., vol. 55, no. 4, pp. 943-950, Apr. 1979.
    [0206] [26] 2, A. F. Habets. [2063, May) Room impulse response (R1R) generator. (Online]. Available: http: // home., Tiscali .nl / ehabets / sorrirgnnsrator. Html; see also: http: // web.
    archive.erg / neb / 20120730003147 / http: // home, tiscali.nl/ehabets / laugh generator.html
    权利要求:
    Claims (7)
    [1]
    BulViDICATION
    1. A UOQ} filter to generate an audio output signal, comprising a plurality of samples of the earn base audio output signal in two or. more microphone input signals, oarauterízado by the audio output signal and the two eu plus microphone input signals are represented. in a frequency cap domain, in which each of the plurality of audio output signal masters S is assigned a position that gives time-frequency ((A, rd) of a number of time-frequency positions ((An)), and where the filter (100} comprises:
    a weighting generator (110) being adapted to receive, for each of the plurality of time-frequency positions ((è, n) 1, the arrival direction information of one or more component (s) of the sound of one or more more sound source (s) or the position information of one or more sound source (s), and being adapted to generate
    weighting i.nfnritáçc for each of the plurality of time-frequency positions ((i., n)), depending on the arrival direction information of one or more component (s) of the sound from one or more source (s) sound [a} da. said time-frequency position ({A, n)} or depending on the position information of one or more sound source (s). (s) of said time-frequency position ((r., n) j, s one output signal generator (12Q) to generate the audio output signal, generating, for each of the plurality of tsmpo-frequency ((A, · n)} positions, one of ρ 1 to 11 of samples from aina 1. de 1 a de a vdio, equal to that assigned to the referred time-frequency classification (. {stη)), depending on the weighting information of that time-frequency position ((ir, η)) and depending of an audio input sample, being assigned the aforementioned time-f position ((à, a)), none of the two or more microphone input signals.
    [2]
    2 <in filter (100) according to claim 1, the ararxiradu by the weighting generator 110) is adapted to generate the weighting information for each of the plurality of the frequency-top positions ((á, n)), depending on the statistical information on the signal or noise components of two or more microphone input signals, depending on the arrival direction information of one or more sound sources (s) gives that position of the soundboard ((! ', - aj) or. depending on the position information of one or more sound source (s) of the said time-freq position ((è, n)) <.
    [3]
    The filter (100) according to claim 2, characterized by the weighting generator (11.0) being adapted to generate the weighting information for each of the plurality of time-customer positions ((k, n)), dapsndcndo of the statistical information about the signal or noise components of two or more signals from the microphone input, in which the statistical information is a power of a noise component, a diffuse signal information, a signal-to-noise information , a diffuse-to-noise information, uses the power of a component of the signal, a power of a diffuse component or a matrix of spectral denatity of the power of a component of the signal, of a noise component or of a diffusion component two or more at the microphone input <
    [4]
    4. A filter (glue) according to the claim characterized by 1 the weighting generator (.110) being adapted to generate the weighting information for each of the plurality of time-frequency positions ((k, Pu depending on the first information indicative of noise information about the primary noise components of two or more microphone input signals and depending on the second information indicative of noise information about the secondary raid components of two or more microphone input signals.
    [5]
    5, The filter (100) according to claim f, characterized in that the weighting generator (110) is adapted to generate the weighting information for each of the plurality of time positions (frequency ((k # n)), depending on the first indicative noise information information on the primary noise components of two or more microphone input signals and depending on the second indicative noise information information on the secondary noise components of two or more microphone input signals, in which the weighting generator (110) is configured to generate the first noise information using statistical information, and in which the weighting generator (IIO) is configured to generate the second noise information without using statistical information, where the statistical information is pradefluids.
    β. a filter (100) according to claim 4 cs 5, maintained by the savings generator (110) to be adapted to generate the weighting information for nothing. one of the plurality of time positions * fr equene ne .1 a. ( : (i, «)), depending on the first noise information on the primary noise components of two or more microphone input signals and depending on the second noise information on the secondary noise components of two or more signals microphone input, where the weighting generator (110) is adapted to generate the weighting information for each of the plurality of time-frequency positions <(i, n)) depending on the formulas:
    Φ, 7 Χ Α [aM ^ a] 'Vg X Λ compreandendu by 4 ; i - Φ ,; and 4q in. which is the first noise information, send a first matrix indicating a first spectral density matrix of. power of primary noise components of one or more microphone input signals, where the second noise information is a second, material indicating a second, spectral density matter, the power of the secondary noise components of one or more microphone input signals , where Á indicates the direction information of s · i i '(s · ·! (sS j,' in cue and a vector indicating the weighting information, where g (L. η) | pjU n )] .... Gt í p / (M n)] j fpfe | cg (A nú,,>
    in. 4 that 'Λ - and primerra runçao o diratividade p.redétinida with complete or actual value depending on the value chagada direction information, and' :: 7S: - ': 7: 77: 7 ......: - -7777. : 7 :: · 7 (IjÀ *; On Í ft h::, :: '::. 777' 7 ' : ' .77.7- „v:.: ... 7: K :, 7.
    where 4 '; and a directional direction of the final depot defined with an area value of 1 or an additional value of eomp1 depending on the arrival direction information.
    [6]
    7. A filter (100) according to any of the 4 to 6 vintage versions, each racier generated by the donor of the computer (110) to be configured to determine the first noise information depending on one or more occurrences between skins minus some of the primary care components of one or more microphone input signals, where one or more consistencies are predefined>
    [7]
    8. A filter (100) according to cc® any one of claims 4 to 7, characterized in that the weighting generator (110) is configured to determine the first noise information depending on a coherence matrix r <i (k) indicating consistencies resulting from the .de., .noise, primary components of two or more microphone input signals, in. that the coherence matrix r ; i (k) is predicted.
    A filter according to

    s and j a c u.mp r i da, an ÂHU indicates direction information


    weighting information depending on the arrival direction information indicating a landing direction of one or more plane waves.
    15x A filter (100) according to any of the previous 1 nd lootations, characterized by the weighting generator (110) comprising a transfer function selection module (104) to force a predefined transfer function and that the weighting generator (1'10) can be configured to generate & Weighting information depending on the direction of arrival information and depending on the transfer function to the challenge.
    16 x a filter (10o) of the color of claim 15, used by the transfer function selection module {104 be configured to provide the predefined transfer function., so that the pridofluid transfer function indies an arbitrary withdrawal pattern depending on the direction of arrival information, for fear that the predefined transfer function indexes a speaker gain depending on arrival direction information, or the way the predefined transfer function indicates a transfer function related to the head depending on the arrival direction information.
    17, Pm. method for generating an audio output signal, comprising a plurality of samples of the audio output signal based on dole or signal input from the microphone, characterized by the signal, from the audio output by
    类似技术:
    公开号 | 公开日 | 专利标题
    BR112015014380A2|2020-01-28|FILTER AND METHOD FOR INFORMED SPACE FILTERING USING MULTIPLE ESTIMATES OF INSTANTARY ARRIVAL DIRECTION.
    Hadad et al.2016|The binaural LCMV beamformer and its performance analysis
    ES2742853T3|2020-02-17|Apparatus and procedure for the direct-environmental decomposition of multichannel for the processing of audio signals
    US9584909B2|2017-02-28|Distributed beamforming based on message passing
    Cornelis et al.2010|Performance analysis of multichannel Wiener filter-based noise reduction in hearing aids under second order statistics estimation errors
    CN106537501B|2019-11-08|Reverberation estimator
    US9564144B2|2017-02-07|System and method for multichannel on-line unsupervised bayesian spectral filtering of real-world acoustic noise
    CN105530580A|2016-04-27|Hearing system
    KR102064902B1|2020-01-10|Globally optimized least squares post filtering for speech enhancement
    Braun et al.2015|A multichannel diffuse power estimator for dereverberation in the presence of multiple sources
    Koutrouvelis et al.2018|A low-cost robust distributed linearly constrained beamformer for wireless acoustic sensor networks with arbitrary topology
    Markovich-Golan et al.2013|Performance of the SDW-MWF with randomly located microphones in a reverberant enclosure
    Beit-On et al.2018|Speaker localization using the direct-path dominance test for arbitrary arrays
    Deleforge et al.2012|A latently constrained mixture model for audio source separation and localization
    Oreinos et al.2014|Validation of realistic acoustic environments for listening tests using directional hearing aids
    Martinez et al.2015|A robust region-based near-field beamformer
    Farmani et al.2018|Sound source localization for hearing aid applications using wireless microphones
    Modhave et al.2016|Design of matrix Wiener filter for noise reduction and speech enhancement in hearing aids
    US11120814B2|2021-09-14|Multi-microphone signal enhancement
    Lawin-Ore et al.2015|Analysis of the average performance of the multi-channel Wiener filter for distributed microphone arrays using statistical room acoustics
    Farmani et al.2018|Bias-compensated informed sound source localization using relative transfer functions
    Arweiler et al.2013|The influence of masker type on early reflection processing and speech intelligibility |
    Thomas et al.2012|Beamformer design using measured microphone directivity patterns: Robustness to modelling error
    Oreinos et al.2013|Objective analysis of higher-order Ambisonics sound-field reproduction for hearing aid applications
    Kurz et al.2017|Prediction of the listening area based on the energy vector
    同族专利:
    公开号 | 公开日
    EP2936830A1|2015-10-28|
    JP2016506664A|2016-03-03|
    JP6196320B2|2017-09-13|
    US10331396B2|2019-06-25|
    RU2015129784A|2017-01-27|
    CN105165026B|2019-08-13|
    EP2936830B1|2016-10-05|
    RU2641319C2|2018-01-17|
    US20150286459A1|2015-10-08|
    EP2747451A1|2014-06-25|
    WO2014095250A1|2014-06-26|
    CN105165026A|2015-12-16|
    EP2936830B8|2017-01-25|
    ES2612528T3|2017-05-17|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    FR2793629B1|1999-05-12|2001-08-03|Matra Nortel Communications|METHOD AND DEVICE FOR CANCELING STEREOPHONIC ECHO WITH FILTERING IN THE FREQUENTIAL DOMAIN|
    CA2354858A1|2001-08-08|2003-02-08|Dspfactory Ltd.|Subband directional audio signal processing using an oversampled filterbank|
    US8098844B2|2002-02-05|2012-01-17|Mh Acoustics, Llc|Dual-microphone spatial noise suppression|
    CA2399159A1|2002-08-16|2004-02-16|Dspfactory Ltd.|Convergence improvement for oversampled subband adaptive filters|
    JP4247037B2|2003-01-29|2009-04-02|株式会社東芝|Audio signal processing method, apparatus and program|
    EP1524879B1|2003-06-30|2014-05-07|Nuance Communications, Inc.|Handsfree system for use in a vehicle|
    EP1538867B1|2003-06-30|2012-07-18|Nuance Communications, Inc.|Handsfree system for use in a vehicle|
    US20070050441A1|2005-08-26|2007-03-01|Step Communications Corporation,A Nevada Corporati|Method and apparatus for improving noise discrimination using attenuation factor|
    US20080004729A1|2006-06-30|2008-01-03|Nokia Corporation|Direct encoding into a directional audio coding format|
    DE602006006664D1|2006-07-10|2009-06-18|Harman Becker Automotive Sys|Reduction of background noise in hands-free systems|
    RS49875B|2006-10-04|2008-08-07|Micronasnit,|System and technique for hands-free voice communication using microphone array|
    US8180062B2|2007-05-30|2012-05-15|Nokia Corporation|Spatial sound zooming|
    US20080298610A1|2007-05-30|2008-12-04|Nokia Corporation|Parameter Space Re-Panning for Spatial Audio|
    JP4455614B2|2007-06-13|2010-04-21|株式会社東芝|Acoustic signal processing method and apparatus|
    AU2008309951B8|2007-10-09|2011-12-22|Dolby International Ab|Method and apparatus for generating a binaural audio signal|
    US8457328B2|2008-04-22|2013-06-04|Nokia Corporation|Method, apparatus and computer program product for utilizing spatial information for audio signal enhancement in a distributed network environment|
    US8724829B2|2008-10-24|2014-05-13|Qualcomm Incorporated|Systems, methods, apparatus, and computer-readable media for coherence detection|
    CN102257401B|2008-12-16|2014-04-02|皇家飞利浦电子股份有限公司|Estimating a sound source location using particle filtering|
    EP2375779A3|2010-03-31|2012-01-18|Fraunhofer-Gesellschaft zur Förderung der Angewandten Forschung e.V.|Apparatus and method for measuring a plurality of loudspeakers and microphone array|
    CN102859591B|2010-04-12|2015-02-18|瑞典爱立信有限公司|Method and arrangement for noise cancellation in a speech encoder|
    US8639499B2|2010-07-28|2014-01-28|Motorola Solutions, Inc.|Formant aided noise cancellation using multiple microphones|
    MX2013006068A|2010-12-03|2013-12-02|Fraunhofer Ges Forschung|Sound acquisition via the extraction of geometrical information from direction of arrival estimates.|
    CN103688522B|2011-05-18|2015-11-25|谷歌公司|Clock drift compensation method and apparatus|US9788128B2|2013-06-14|2017-10-10|Gn Hearing A/S|Hearing instrument with off-line speech messages|
    GB2521649B|2013-12-27|2018-12-12|Nokia Technologies Oy|Method, apparatus, computer program code and storage medium for processing audio signals|
    EP2975609A1|2014-07-15|2016-01-20|Ecole Polytechnique Federale De Lausanne |Optimal acoustic rake receiver|
    EP3206415B1|2014-10-10|2019-09-04|Sony Corporation|Sound processing device, method, and program|
    US10242690B2|2014-12-12|2019-03-26|Nuance Communications, Inc.|System and method for speech enhancement using a coherent to diffuse sound ratio|
    EP3220386A1|2016-03-18|2017-09-20|Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V.|Apparatus and method for harmonic-percussive-residual sound separation using a structure tensor on spectrograms|
    RU2714579C1|2016-03-18|2020-02-18|Фраунхофер-Гезелльшафт Цур Фердерунг Дер Ангевандтен Форшунг Е.Ф.|Apparatus and method of reconstructing phase information using structural tensor on spectrograms|
    CN106028227B|2016-07-08|2019-05-24|乐鑫信息科技（上海）股份有限公司|Distributed microphone array and its applicable sonic location system|
    CN106060743A|2016-08-03|2016-10-26|上海山景集成电路股份有限公司|Microphone, microphone combination and microphone signal processing method|
    CN106569773A|2016-10-31|2017-04-19|努比亚技术有限公司|Terminal and voice interaction processing method|
    CN106782590B|2016-12-14|2020-10-09|南京信息工程大学|Microphone array beam forming method based on reverberation environment|
    GB2563952A|2017-06-29|2019-01-02|Cirrus Logic Int Semiconductor Ltd|Speaker identification|
    US10679617B2|2017-12-06|2020-06-09|Synaptics Incorporated|Voice enhancement in audio signals through modified generalized eigenvalue beamformer|
    TWI690218B|2018-06-15|2020-04-01|瑞昱半導體股份有限公司|headset|
    CN110636400B|2018-06-25|2021-03-16|瑞昱半导体股份有限公司|Earphone set|
    US20200058317A1|2018-08-14|2020-02-20|Bose Corporation|Playback enhancement in audio systems|
    CN109839612B|2018-08-31|2022-03-01|大象声科（深圳）科技有限公司|Sound source direction estimation method and device based on time-frequency masking and deep neural network|
    CN109286875B|2018-09-29|2021-01-01|百度在线网络技术（北京）有限公司|Method, apparatus, electronic device and storage medium for directional sound pickup|
    US20200184994A1|2018-12-07|2020-06-11|Nuance Communications, Inc.|System and method for acoustic localization of multiple sources using spatial pre-filtering|
    CN111175693A|2020-01-19|2020-05-19|河北科技大学|Direction-of-arrival estimation method and direction-of-arrival estimation device|
    CN113203987A|2021-07-05|2021-08-03|成都启英泰伦科技有限公司|Multi-sound-source direction estimation method based on K-means clustering|
    法律状态:
    2018-11-21| B06F| Objections, documents and/or translations needed after an examination request according [chapter 6.6 patent gazette]|
    2020-02-04| B15I| Others concerning applications: loss of priority|Free format text: VIDE E-PARECER |
    2020-03-24| B152| Others concerning applications: decision cancelled [chapter 15.32 patent gazette]|Free format text: ANULADA A PUBLICACAO CODIGO 15.9 NA RPI NO 2561 DE 04/02/2020, TENDO EM VISTA O DOCUMENTO DE CESSAO APRESENTADO NA PETICAO NO 018150007962. CUMPRE FRISAR QUE, DE ACORDO COM A PETICAO NO 870200024187, A CITADA PETICAO NAO ESTAVA ASSOCIADA CORRETAMENTE AO PEDIDO POR ERRO DO USUARIO NO MOMENTO DO EXAME DE ADMISSIBILIDADE, O QUE LEVOU A DECRETACAO DA PERDA DA PRIORIDADE AGORA ANULADA. |
    2020-04-14| B06U| Preliminary requirement: requests with searches performed by other patent offices: procedure suspended [chapter 6.21 patent gazette]|
    2021-10-13| B350| Update of information on the portal [chapter 15.35 patent gazette]|
    优先权:
    申请号 | 申请日 | 专利标题
    US201261740866P| true| 2012-12-21|2012-12-21|
    EP13169163.6A|EP2747451A1|2012-12-21|2013-05-24|Filter and method for informed spatial filtering using multiple instantaneous direction-of-arrivial estimates|
    PCT/EP2013/074650|WO2014095250A1|2012-12-21|2013-11-25|Filter and method for informed spatial filtering using multiple instantaneous direction-of-arrivial estimates|
    [返回顶部]