DIRECTION OF ARRIVAL ESTIMATION IN MINIATURE DEVICES USING A SOUND SENSOR ARRAY

Abstract

A hearing device comprises a sound system for estimating the direction of arrival of sound emitted by one or more sound sources creating a sound field. The sound system comprises an array of N sound receiving transducers (microphones), each providing an electric input signal, a processing unit comprising a) a model unit comprising a parametric model configured to be able to describe the sound field at the array as a function of the direction of arrival in a region surrounding and adjacent to the array; b) a model optimizing unit configured to optimize said model with respect to its parameters based on said sound samples; c) a cost optimizing unit configured to minimize a cost function of the model with respect to said direction of arrivals; d) an estimating unit configured to estimate the direction of arrival based on said parametric model with the optimized parameters and the optimized cost function.

Claims

1. A hearing device, e.g. a hearing aid, comprising A sound system for estimating the direction of arrival of a sound signal emitted from one or more sound sources, the system comprising: a sound sensor unit comprising an array of N sound receiving transducers, each providing an electric input signal; a sampling unit for providing at least one sample of said surrounding sound field from each of said electric input signals at substantially the same time instant; and a processing unit comprising a model unit comprising a parametric model configured to be able to describe the sound field at the array as a function of the direction of arrival in a region surrounding and adjacent to the array; a model optimizing unit configured to optimize said model with respect to its parameters based on said sound samples; a cost optimizing unit configured to minimize a cost function of the model with respect to said direction of arrivals; an estimating unit configured to estimate the direction of arrival based on said parametric model with the optimized parameters and the optimized cost function.

2. A hearing device according to claim 1, wherein said array of sound sensitive sensors is or comprises an array of microphones configured to be worn by a user, e.g. mounted on the head of a user.

3. A hearing device according to claim 1 wherein said array of sound sensitive sensors comprises microphones integrated on a chip, or MEMS microphones.

4. A hearing device according to claim 1 wherein the number N of sound sensitive sensors is larger than 3, e.g. in the range from 3 to 10 or larger than 10, such as larger than 50.

5. A hearing device according to claim 1 configured to provide that the distance between adjacent sound sensitive sensors is smaller than 0.01 m, e.g. smaller than or equal to 0.007 m.

6. A hearing device according to claim 1 comprising a beamformer filtering unit for providing a beamformed signal based on one or more beamformer input signals representing said surrounding sound field, and configured to use said estimate the direction of arrival in the determination of the beamformed signal.

7. A hearing device according to claim 6 configured to provide that said beamformer filtering unit receives at least some of said electric input signals from said sound sensor array, and provides said beamformed signal as a weighted combination of said at least some of said electric input signals and possibly further electric input signals.

8. A hearing device according to claim 1 comprising filter, e.g. a low pass filter, configured to isolate a fundamental frequency of the electric input signals corresponding to a human voice from higher lying harmonics and providing respective filtered electric input signals, and to use such filtered electric input signals for identification of the direction of arrival.

9. A hearing device according to claim 1 comprising a tracking unit configured to track one or more sound sources over time.

10. A hearing device according to claim 1 configured to estimate a direction of arrival to a multitude of active target sound sources.

11. A hearing device according to claim 10 when dependent on claim 6 wherein the beamformer filtering unit is configured to determine a beam pattern based on estimates of direction of arrival of a multitude of active target sound sources.

12. A hearing device according to claim 11 configured to activate beams at different direction of arrivals at the same time.

13. A hearing device according to claim 1 comprising a network of arrays of sound sensitive sensors.

14. A hearing device according the claim 1 comprising a hearing aid.

15. A hearing device according to claim 14 wherein the hearing aid comprises a behind-the-ear hearing aid or an in-the-ear hearing aid or hearing glasses or a headband.

16. A hearing device according to claim 1 comprising a user interface allowing a user to control functionality of the device.

17. A method for estimating the direction of arrival of sound emitted by one or more sound sources creating a sound field in the surroundings of the one or more sound sources, the method comprising: providing an array of sound sensitive sensors, each sensor providing an electric input signal comprising a stream of digital samples representing the sound field of the surroundings; providing at least one sample of said surrounding sound field from each of said sensors at substantially the same time instant; providing a parametric model configured to be able to describe the sound field as a function of the direction of arrival in a region surrounding and adjacent to said array; optimizing said model with respect to its parameters based on said sound samples; and minimizing a cost function of the model with respect to said direction of arrivals.

18. A method according to claim 17, wherein said model is a Taylor expansion model and the sound signal received at each of said respective sensor is decomposed in a Taylor series expansion of order L, in which the signal derivatives are determined by a linear least squares method, and wherein the direction of arrival is estimated by the expression: $\hat{}=\underset{}{\mathrm{argmin}}.Math.V\left(,\hat{S}\left(\right)\right)=\underset{}{\mathrm{argmin}}.Math.{.Math.z-\left(\right).Math.\hat{S}\left(\right).Math.}^2$ where .sup.2 represents the least squares loss (or error) function, z is a mixed signal comprising sound signal s and additional noise , and T() is a vector comprising the first L Taylor expansion coefficients.

19. A data processing system comprising a processor and program code means for causing the processor to perform the method of claim 17.

20. A non-transitory application, termed an APP, comprising executable instructions configured to be executed on an auxiliary device to implement a user interface for a hearing device as described in claim 1.

21. An APP according to claim 20 configured to run on cellular phone, e.g. a smartphone, or on another portable device allowing communication with said hearing device.

Description

BRIEF DESCRIPTION OF DRAWINGS

[0082] The aspects of the disclosure may be best understood from the following detailed description taken in conjunction with the accompanying figures. The figures are schematic and simplified for clarity, and they just show details to improve the understanding of the claims, while other details are left out. Throughout, the same reference numerals are used for identical or corresponding parts. The individual features of each aspect may each be combined with any or all features of the other aspects. These and other aspects, features and/or technical effect will be apparent from and elucidated with reference to the illustrations described hereinafter in which:

[0083] FIG. 1A shows a schematic plane view of a microphone array that can be used in embodiments of the present disclosure;

[0084] FIG. 1B shows an example of output voltages as a function of time from the four microphones in the array shown in FIG. 1A;

[0085] FIG. 2A shows an actual and two hypothetical propagating sound waves passing a microphone array comprising four microphones in different directions relative to the array; and

[0086] FIG. 2B shows the amplitude of the response from the microphone array in FIG. 2A corresponding to each respective of passing sound waves shown in FIG. 2A;

[0087] FIG. 3A shows a user wearing left and right hearing aids in a listening situation comprising a target sound source and noise, and

[0088] FIG. 3B shows a user wearing left and right hearing aids and a headband comprising a microphone array;

[0089] FIG. 4 shows an embodiment of a hearing aid according to the present disclosure comprising a BTE-part located behind an ear or a user and an ITE part located in an ear canal of the user, wherein the hearing aid is in communication with an auxiliary device comprising a user interface for the hearing aid;

[0090] FIG. 5 shows a flow diagram for a method of providing a direction of arrival of sound from a sound source to a sound system according to the present disclosure;

[0091] FIG. 6 shows a simplified block diagram of an embodiment of a sound system according to the present disclosure;

[0092] FIG. 7 shows a simplified block diagram of an embodiment of a hearing device comprising a beamformer filtering unit and a sound system according to the present disclosure;

[0093] FIG. 8 shows an application of a hearing device or a hearing system according to the present disclosure for segregating individual sound sources in a multi-sound source environment;

[0094] FIG. 9 schematically shows a plot of attenuation versus angle (at a specific frequency) for a sound system comprising an array of sound sensitive sensors according to the present disclosure; and

[0095] FIG. 10 illustrates a multi-target sound source direction of arrival estimation situation where a hearing system according to the present disclosure estimates a target maintaining beam pattern having focus on all (three) target sound sources.

DETAILED DESCRIPTION

[0096] The detailed description set forth below in connection with the appended drawings is intended as a description of non-limiting example embodiments of the method and system according to the present disclosure.

[0097] With reference to FIG. 1A there is shown a schematic plane view of a microphone array that can be used in embodiments of the present disclosure, but it is understood that many other configurations of arrays could also be used without departing from the scope of the present disclosure. The array comprises four microphones 1, 2, 3 and 4. FIG. 1A also shows a possible directional characteristic obtainable with the array, where the directional characteristic comprises a main lobe 5, two side lobes 6 and 8 and a rear lobe 7. Many other directional characteristics can be obtained dependent inter alia on the specific signal processing carried out on each specific microphone signal.

[0098] With reference to FIG. 1B there is shown an example of output voltages (between 1 V and +1 V) as a function of time (from 0 to 15 ms) from the four microphones in the array shown in FIG. 1A generated when a sound wave propagates through the array in a specific direction relative to the array. The respective output voltages are designated by z.sub.1, z.sub.2, z.sub.3 and z.sub.4, which is the same notation that will be used in the detailed description given below. In the example shown in FIG. 1B, the sound wave impinges on microphone 1 first followed by microphones 4, 2 and 3, which gives rise to the relative delays of the output voltages shown in FIG. 1B.

[0099] With reference to FIG. 2A there is shown an actual propagating sound wave 9 travelling in the direction indicated by arrow 10 and two hypothetical propagating sound waves 11 and 13 travelling in directions 12 and 14, respectively. In this example, the microphone array comprises four microphones 15, 16, 17, 18 and one of these microphonesin the shown example microphone 18can be regarded as a reference microphone R positioned at the origin of the coordinate system x, y. The direction of propagation is uniquely indicated by the angle in this two-dimensional example, but propagation in a three-dimensional space would also be possible within the scope of the disclosure and would require two angles , to uniquely indicate the direction of propagation of the sound wave (cf. also top part of FIG. 3A). The direction-of-arrival (DOA) for the acoustic waveform (9) is captured by the microphone array (stars 15, 16, 17, 18 indicating the individual microphones in FIG. 2A). FIG. 2A shows three (out of an infinite number of) hypotheses for the incoming directions (10, 12, 14). Only one of the directions (10) gives a reasonable and smooth reconstruction in the time domain, and thus defines the DOA. FIG. 2A can be taken to illustrate the scenario in the space domain (as indicated in the top part of FIG. 2A).

[0100] With reference to FIG. 2B there is shown the amplitude of the response (the output voltage of the complete microphone array after suitable processing of the respective microphone signals versus time) from the four microphones of the array in FIG. 2A corresponding to each respective of the sound waves shown in FIG. 2A (the solid dots corresponding to wave 9 and direction 10, the cross-hatched dots corresponding to wave 11 and direction 12, and the vertically hatched dots corresponding to wave 13 and direction 14, respectively).

[0101] FIG. 2B illustrates, in an exemplified form, the basic concept of the present disclosure. The amplitude of the output from the array as a function of time resulting from sound wave 9 passing the array is shown at four specific points in time by the solid dots 19 in FIG. 2B. Similarly, the passing of sound wave 11 through the array will result in the output indicated by the dots 20 and the passing of sound wave 13 will result in the output indicated by the dots 21 in FIG. 2B. As it appears, only one of the sound waves (wave 9, direction 10), i.e. one of the infinitely many directions of propagation of the sound wave, gives a reasonable and smooth reconstruction in the time domain, and thus defines the direction of arrival (DOA) estimated according to the principles of the present disclosure. FIG. 2A can be taken to illustrate the scenario in the time domain (as indicated in the top part of FIG. 2B).

[0102] The four microphones in FIG. 2A can be seen as a subset of an array, e.g. of a microphone array, e.g. implemented as a headband filled with microphones all around the head. The proposed method selects the wave front (9) in FIG. 2A passing the microphones in a smooth way represented in FIG. 2B by solid dots lying on amplitude versus time curve 22. This encompasses the most probable direction of arrival (DOA), with angle . If the angle is just a little bit smaller or larger, the pattern in FIG. 2B would not be as smooth and thus less probable as DOA. Thus by listening in the direction of (i.e. by directing a beam in the direction of using an appropriate combination of the microphone signals of the array, i.e. providing a beamformer filtering unit) a sensitive directional microphone array pointing in the direction may be implemented.

[0103] In the following a specific embodiment of far-field direction of arrival (DOA) estimation according to the present disclosure will be described.

A. Signal Model

[0104] Assume a plane propagation model where the sensors are located at (x.sub.n, y.sub.n) for n=1, 2, . . . , N. In this section it is assumed that the sensors are much closer to the origin of the coordinate system than the sound sources (e.g. so that plane wave approximation is valid).

[0105] A sound source emits an acoustic signal s(t) which reaches sensor n with a delay and additive Gaussian noise given by the expression:

.sub.n(t)(0,.sub..sub.n.sup.2)

[0106] The signal that reaches sensor n is then given by the expression:

z.sub.n(t)=s(t+.sub.n())+.sub.n(t),(1)

which is one of the standard assumptions in DOA literature. In expression (1), is the angle between the x-axis of the coordinate system and the vector between sensor that is defined as a reference sensor R and the specific sound source (c.f. FIG. 2A). The delay can then be expressed as follows:

[00002]$\begin{array}{cc}{}_n\left(\right)=\frac{x_n.Math.\cos .Math..Math.+y_{n.Math.}.Math.\sin .Math..Math.}{c}.& \left(2\right)\end{array}$

where c is the speed of sound (e.g. 343 m/s in air).

B. Assumptions

[0107] The main assumption is that the delay differences (between the microphone signals) are much smaller than the signal variation, which in terms of frequency can be stated as:

[00003]$\begin{array}{cc}.Math.{}_n\left(\right)-{}_m\left(\right).Math.\frac{1}{f_{\mathrm{ma}.Math..Math.x}},m,n.& \left(3\right)\end{array}$

[0108] Since the delay difference is bounded by the array size D (for instance the typical diameter of the human head), expression (3) can be stated as:

[00004]$\begin{array}{cc}D\frac{c}{f_{\mathrm{ma}.Math..Math.x}}.& \left(4\right)\end{array}$

[0109] As an example, a motorized vehicle has the fundamental frequency of sound emission below 100 Hz (6000 rpm). The shortest wavelength is thus about 3 meters, and the array should be much smaller than this, for instance 0.3 meters. In the case of an average human head, D could be approximately 0.25 meters. In another example, fundamental frequencies of the human voice are used for identification of the direction of arrival. A fundamental frequency of a male voice is typically in the range from 85 Hz to 180 Hz. Correspondingly, a fundamental frequency of a female voice typically in the range from 165 Hz to 255 Hz. Hence a head mounted sensor array utilizing the scheme for estimating a direction of arrival according to the present disclosure may be used.

C. Taylor Expansion

[0110] If the condition in (3) is satisfied, the signal varies smoothly over the array, and a Taylor

[00005]$\begin{array}{cc}s\left(t+{}_n\left(\right)\right)=\underset{i=0}{\overset{L}{.Math.}}.Math.\frac{{}_n^i\left(\right)}{i!}.Math.\frac{d^i}{{\mathrm{dt}}^i}.Math.s\left(t\right)+{}_n\left(t\right).& \left(5\right)\end{array}$

series expansion of order L will represent the local behavior of the signal at sensor n. The following expression then applies:

[0111] The above expression can be written as a linear regression:

s(t+.sub.n())=.sub.n.sup.T()s+.sub.n(t)(6)

where .sub.n(t) denotes the higher order terms of the Taylor expansion which in the following will be neglected as a consequence of assumption (3), and wherein:

[00006]$\begin{array}{cc}{}_n^T\left(\right)=\left(\begin{array}{ccccc}1& {}_n\left(\right)& \frac{1}{2!}.Math.{}_n^2\left(\right)& \mathrm{.Math.}& \frac{1}{L!}.Math.{}_n^L\left(\right)\end{array}\right)& \left(7.Math.a\right)\\ S^T\left(t\right)=\left(\begin{array}{ccccc}{s}^{\left(0\right)}& s^{\left(1\right)}& s^{\left(2\right)}& \mathrm{.Math.}& s^{\left(L\right)}\end{array}\right).Math..Math.\mathrm{and}.Math..Math.s^{\left(L\right)}=\frac{d^L}{{\mathrm{dt}}^L}.Math.s\left(t\right).& \left(7.Math.b\right)\end{array}$

[0112] The original model given in expression (1) can thus be written as:

z.sub.n(t)=.sub.n.sup.T()S(t)+.sub.n(t).(8)

[0113] Considering the array, the following set of equations that are linear in the Taylor expansion parameters but nonlinear with respect to , apply:

[00007]$\begin{array}{cc}{z}_1\left(t\right)={}_1^T\left(\right).Math.s\left(t\right)+{}_1.Math..Math.z_2\left(t\right)={}_2^T\left(\right).Math.s\left(t\right)+{}_2.Math..Math.\mathrm{.Math.}.Math..Math.z_N\left(t\right)={}_N^T\left(\right).Math.s\left(t\right)+{}_N& \left(9\right)\end{array}$

[0114] Where N denotes the number sound sensitive sensors (e.g. microphones).

[0115] This set of equations can conveniently be written in matrix form as:

[00008]$\begin{array}{cc}z\left(t\right)=\left(\right).Math.S\left(t\right)+.Math..Math.\mathrm{where}& \left(10\right)\\ \left(\right)=\left(\begin{array}{c}{}_1^T\left(\right)\\ {}_2^T\left(\right)\\ \mathrm{.Math.}\\ {}_N^T\left(\right)\end{array}\right)& \left(11\right)\end{array}$

D. DOA Estimation

[0116] An essential feature of the partially linear model in expression (10) is that the least squares (LS) estimate, which coincides with the maximum likelihood (ML) estimate for Gaussian noise is conveniently computed by searching for the optimal direction of arrival , where all the linear parameters can be estimated analytically. That is, the optimization only concerns a scalar parameter, independently of the number N of array elements or the order L of the Taylor expansion. This also enables parallel structures of implementations, where the DOA angles are gridded and the measure of LS fit can be computed completely in parallel.

[0117] The LS estimate is per definition given by the expression:

[00009]$\begin{array}{cc}\left(\hat{},\hat{S}\right)=\underset{,S}{\mathrm{argmin}}.Math.V\left(,S\right)& \left(12\right)\end{array}$

where V denotes the LS loss function given by:

V(,S)=z()S.sup.2(13)

[0118] The linear sub-structure makes the estimation problem fit the separable least squares (SLS) framework which eventually makes solving the optimization problem more computationally efficient.

[0119] If is a fixed parameter, the estimate of s is given by:

[00010]$\begin{array}{cc}\begin{array}{c}\hat{s}\left(\right)=.Math.\underset{s}{\mathrm{argmin}}.Math.V\left(,s\right)\\ =.Math.\underset{s}{\mathrm{argmin}}.Math.{.Math.z-\left(\right).Math.s.Math.}_F^2\end{array}& \left(14\right)\end{array}$

[0120] For the above optimization problem, the estimate can be computed using least squares:

()=.sup.t()z,(15)

[0121] Where the cross-symbol denotes Moore-Penrose pseudo-inverse.

[0122] This leads to the following estimate of the direction of arrival :

[00011]$\begin{array}{cc}\hat{}=\underset{}{\mathrm{argmin}}.Math.V\left(,\hat{S}\left(\right)\right)=\underset{}{\mathrm{argmin}}.Math.{.Math.z-\left(\right).Math.\hat{S}\left(\right).Math.}^2.& \left(16\right)\end{array}$

[0123] If is one or two dimensional, the estimate of can be computed quite efficiently by evaluating it over a fine grid, which is, as mentioned above, suitable for parallel implementations.

[0124] In the foregoing, only a single sample of the array has been formulated. However, a batch of samples at t=t.sub.1, t.sub.2, . . . , t.sub.m can be modeled in the same way by stacking these samples into an mN vector and expanding T() into a block diagonal matrix with m identical blocks, provided that varies insignificantly over the duration of the batch.

E. Multiple Signal Sources

[0125] In principle, the linearized signal model (10) can be extended to multiple signal sources straightforwardly. If the number of signal sources is denoted by K, expression (10) becomes:

[00012]$\begin{array}{cc}z\left(t\right)=\underset{k=1}{\overset{K}{.Math.}}.Math.\left({}_k\right).Math.S_k\left(t\right)+.& \left(17\right)\end{array}$

[0126] To satisfy the identification criteria, the number of unknowns must be less than the number of observations N.sub.obs, i.e.:

N.sub.obs(L+1)K

corresponding to K sets of Taylor expansions with L unknowns, and one extra angle parameter to each source. Thus, the number of sensor elements in the array increases linearly with the number of signal sources.

F. Design Issues

[0127] Real signals are seldom band-limited, and therefore the maximum frequency f.sub.max will not be well defined. However, a low-pass filter can always be applied to all sensors. Therefore, f.sub.max can be considered a design parameter, just as the size D of the array. The order L of the Taylor expansion can also be freely chosen by the particular use. In one practical implementation of the present disclosure, the design order is as follows:

1) D is given by the specific construction of the array.
2) f.sub.max is selected based on source excitation to get the best possible signal to noise ratio (SNR). However, it is necessary that f.sub.max<<c/D.
3) The Taylor expansion order L is a monotonically increasing function of Df.sub.max and L must satisfy the constraint (L+2) being less than or equal to N in order to obtain a unique solution.

[0128] It may be recommended to start with a first order Taylor expansion, i.e. L=1.

[0129] FIG. 3A schematically illustrates the geometrical setup of a hearing aid user (U) wearing left and right hearing devices, e.g. hearing aids, (L-HD, R-HD) (e.g. forming a binaural hearing aid system) in a listening situation comprising a sound source (S), whichat a given point in timeis located at a specific point in space (e.g. defined by coordinates in a coordinate system, here having its centre in the head of the user U between the ears, as indicated by arrows x, y, z and as illustrated in the coordinate system in the top part of FIG. 3A). The coordinates of the sound source S in a spherical representation are (.sub.s, r.sub.s, (.sub.s=90)) relative to the user. The sound scene further comprises one or more noise sources (Noise), which may be distributed or spatially located. Each of the left and right hearing aids comprises a part, often termed a BTE-part (BTE), adapted for being located behind an ear (Left ear, Right ear) of the user (U). The BTE-part comprises first (Front) and second (Rear) microphones (FM.sub.L, RM.sub.L; FM.sub.R, RM.sub.R) for converting an input sound to first and second electric input signals, respectively. The first and second microphones (FM, RM) of a given BTE-part, when located behind the relevant ear of the user (U), are characterized by transfer functions H.sub.FM(, , r, k) and H.sub.RM(, , r, k) representative of propagation of sound from the sound source S located at (, , r) around the BTE-part to the first and second microphones of the hearing aid (L-HD, R-HD) in question, where k is a frequency index. In the setup of FIG. 3, the target signal source S is assumed to be located in the frontal half-plane relative to the user (U) at an angle (90 in FIG. 3A) to the direction of the nose of the user (cf. LOOK-DIR in FIG. 3A), and to a microphone axis of the BTE-parts (cf. e.g. reference directions REF-DIR.sub.L, REF-DIR.sub.R, of the left and right hearing aids). A vector d comprising the transfer functions d=(H.sub.FM(, , r, k), H.sub.RM(, , r, k)) is termed the look vector for the microphones of the hearing device in question.

[0130] The first and second microphones of a given BTE-part are located at predefined distance L.sub.M apart (often referred to as microphone distance d). The two BTE-parts and thus the respective microphones of the left and right hearing aids, are located a distance a apart, when mounted on the user's head in an operational mode. In a situation where the user U is located in the acoustic far-field from source S, the distance r.sub.s from the user U to the sound source S can be assumed to be much larger than any distance (L.sub.M and/or a) between two neighboring microphones of the microphone array (L.sub.M, a<<r.sub.s), r.sub.s being not to scale as indicated by the broken indication of vector r.sub.s in FIG. 3A). The view in FIG. 3A is a planar view in a horizontal plane through the microphones of the first and second hearing aids (perpendicular to a vertical direction, indicated by out-of-plane arrow VERT-DIR (z) in FIG. 3A) and corresponding to plane z=0 (=90). In a simplified model, it can be assumed that the sound sources (S.sub.i) are located in a horizontal plane (e.g. the one shown in FIG. 3A). An interaural communication link between the left and right hearing aids is indicated by dashed arrow (denoted IAL) in FIG. 3A. The interaural communication link (IAL) is e.g. configured to allow audio signals (or parts thereof, e.g. selected frequency ranges, e.g. a low frequency part) and control and/or status signals to be exchanged between the hearing aids (or forwarded from one to the other or to an auxiliary device). In an embodiment, all four microphone signals (or selected frequency ranges thereof) are available in one or both hearing devices (and/or in an auxiliary device, cf. e.g. FIG. 4). In an embodiment, the four microphones (FM.sub.L, RM.sub.L, FM.sub.R, RM.sub.R) form part of (such as constitute) a sensor array in the sense of the present disclosure. In an embodiment, an array of microphones (e.g. each comprising three or more microphones, e.g. four microphones as shown in FIG. 1A) are located at each ear (and/or between the ears, e.g. on the forehead or distributed around the head, e.g. in a headband, cf. e.g. FIG. 3B, and networked together via the interaural link IAL (or respective communication links to an auxiliary device, cf. e.g. FIG. 4, or wired connections). A direction of arrival (DOA) of the sound from the sound source S to the user (U)in practice to the mid-point of the user's headis indicated in FIGS. 3A and 3B. The use of reference coordinate system is a matter of choice for the skilled person. An exemplary spherical coordinate system is shown for reference in the top part of FIG. 3A. It is assumed that the location of microphones of the microphone array or the microphone arrays are known (or can be determined) in a common reference coordinate system. This is indicated by the vector r.sub.RML to and coordinates of rear microphone RM.sub.L of the left hearing device L-HD (r.sub.RML=(x.sub.RML, y.sub.RML, z.sub.RML) (r.sub.RML, .sub.RML, .sub.RML)). This and corresponding coordinates of the other microphones are e.g. stored in a memory of the sound system.

[0131] FIG. 3B illustrates a scenario as in FIG. 3A, but where the sound system additionally comprises an array of microphones located on the head (here the forehead) of the user (here indicated to form part of a headband (Headband). The locations of the (M) individual microphones HBM.sub.1, . . . , HBM.sub.M of the headband are assumed to be known, here having equal distance L.sub.HBM between neighboring units. The microphones of the headband are connected to a processor of (at least) one of the hearing devices or to a separate processor, e.g. located in an auxiliary device, being configured to process all electric input signals from the individual microphones of the headband and of the left and right hearing devices. The microphones (e.g. being more than the two shown in FIGS. 3A and 3B, e.g. three or four) of the left and right hearing devices and the microphones of the headband may be considered as three individual arrays of microphones tied together by a network. As in the scenario of FIG. 3A it is assumed that the locations (e.g. coordinates (x, y) or (r, )) of the individual microphones of the headband are known, and e.g. stored in a memory of the sound system. In the embodiment of FIG. 3B, the array of microphones is located over a limited length (area) of the headband. The microphones may, however, otherwise be distributed around the headband, e.g. evenly, or in individual groups, or be non-periodicly or randomly distributed (with known locations), e.g. with higher density (or a larger number) towards the front than towards the rear. Thereby, all angles around a user wearing the headband may be attended to.

[0132] FIG. 4 illustrates an exemplary hearing aid (HD) formed as a receiver in the ear (RITE) type hearing aid comprising a BTE-part (BTE) adapted for being located behind pinna and a part (ITE) comprising an output transducer (e.g. a loudspeaker/receiver, SPK) adapted for being located in an ear canal (Ear canal) of the user. The BTE-part (BTE) and the ITE-part (ITE) are connected (e.g. electrically connected) by a connecting element (/C). In the embodiment of a hearing aid of FIG. 4, the BTE part (BTE) comprises two input transducers (here microphones) (FM, RM) each for providing an electric input audio signal representative of an input sound signal (S.sub.BTE) from the environment. In the scenario of FIG. 4, the input sound signal S.sub.BTE includes a contribution from sound source S, S being e.g. sufficiently far away from the user (and thus from hearing device HD) so that its contribution to the acoustic signal S.sub.BTE is in the acoustic far-field (alternatively, S may be located in the near-field relative to the hearing aid microphones). The hearing aid of FIG. 4 further comprises two wireless receivers (WLR.sub.1, WLR.sub.2) for providing respective directly received auxiliary audio and/or information signals. The hearing aid (HD) further comprises a substrate (SUB) whereon a number of electronic components are mounted, functionally partitioned according to the application in question (analogue, digital, passive components, etc.), but including a configurable signal processing unit (SPU), a beamformer filtering unit (BFU), and a memory unit (MEM) coupled to each other and to input and output units via electrical conductors Wx. The mentioned functional units (as well as other components) may be partitioned in circuits and components according to the application in question (e.g. with a view to size, power consumption, analogue vs. digital processing, etc.), e.g. integrated in one or more integrated circuits, or as a combination of one or more integrated circuits and one or more separate electronic components (e.g. inductor, capacitor, etc.). The configurable signal processing unit (SPU) provides an enhanced audio signal, which is intended to be presented to a user. In the embodiment of a hearing aid device in FIG. 4, the ITE part (ITE) comprises an output unit in the form of a loudspeaker (receiver) (SPK) for converting the electric signal (OUT) to an acoustic signal (providing, or contributing to, acoustic signal S.sub.ED at the ear drum (Ear drum). In an embodiment, the ITE-part further comprises an input unit comprising an input transducer (e.g. a microphone) (M.sub.ITE) for providing an electric input audio signal representative of an input sound signal S.sub.ITE from the environment (including from sound source S) at or in the ear canal. In another embodiment, the hearing aid may comprise only the BTE-microphones (FM, RM). In another embodiment, the hearing aid may comprise more than the three microphones (FM, RM, M.sub.ITE). In yet another embodiment, the hearing aid may comprise an input unit (IT.sub.3) located elsewhere than at the ear canal in combination with one or more input units located in the BTE-part and/or the ITE-part. The ITE-part further comprises a guiding element, e.g. a dome, (DO) for guiding and positioning the ITE-part in the ear canal of the user.

[0133] The hearing aid (HD) exemplified in FIG. 4 is a portable device and further comprises a battery, e.g. a rechargeable battery, (BAT) for energizing electronic components of the BTE- and ITE-parts.

[0134] The hearing aid (HD) may e.g. comprise a directional microphone system (beamformer filtering unit (BFU)) adapted to enhance a target acoustic source among a multitude of acoustic sources in the local environment of the user wearing the hearing aid device. In an embodiment, the directional system is adapted to detect (such as adaptively detect) from which direction a particular part of the microphone signal (e.g. a target part and/or a noise part) originates (the DOA), as taught by the present disclosure. In an embodiment, the beamformer filtering unit is adapted to receive inputs from a user interface (e.g. a remote control or a smartphone) regarding an estimate of the present target direction (DOA). The memory unit (MEM) may e.g. comprise predefined (or adaptively determined) complex, frequency dependent constants (W.sub.ij) defining predefined (or adaptively determined) or fixed beam patterns (e.g. omni-directional, target cancelling, etc.), together defining a beamformed signal. The memory MEM may further have stored values of the coordinates of the individual microphones (FM, RM, M.sub.ITE, . . . ) of the hearing device in an appropriate coordinate system (at least relative to a reference point, e.g. a fixed location in the hearing device).

[0135] The hearing aid of FIG. 4 may constitute or form part of a hearing aid and/or a binaural hearing aid system according to the present disclosure. The hearing aid may comprise an analysis filter bank, and the processing of an audio signal in a forward path of the hearing aid may e.g. be performed fully or partially in the time-frequency domain. Likewise, the processing of signals in an analysis or control path of the hearing aid may be fully or partially performed in the time-frequency domain.

[0136] The hearing aid (HD) according to the present disclosure may comprise a user interface UI, e.g. as shown in the lower part of FIG. 4 implemented in an auxiliary device (AUX), e.g. a remote control, e.g. implemented as an APP in a smartphone or other portable (or stationary) electronic device. In the embodiment of FIG. 4, the screen of the user interface (UI) illustrates a Target direction APP. The display of the auxiliary device schematically illustrates a screen of the Direction-of-Arrival-APP instructing a user to Indicate direction to target sound source S. A (approximate) direction to the present target sound source (S) may be indicated from the user interface, e.g. by dragging the sound source symbol to a currently relevant direction relative to the user. The currently selected target direction is to the left of the frontal direction (at 45 (+90), cf. FIG. 3) as indicated by the bold arrow to the sound source S. The auxiliary device and the hearing aid are adapted to allow communication of data representative of the currently indicated direction (e.g. for use as a first estimate for the algorithm) to the hearing aid via a, e.g. wireless, communication link (cf. dashed arrow WL2 in FIG. 4). The auxiliary device (AUX) and the hearing aid (HD) are adapted to allow the exchange of data representative of a direction to the target sound source (DOA) (and optionally audio signals and other control or information signals) between them via a, e.g. wireless, communication link (cf. dashed arrow WL2 in FIG. 4). The communication link WL2 may e.g. be based on far field communication, e.g. Bluetooth or Bluetooth Low Energy (or similar technology), implemented by appropriate antenna and transceiver circuitry in the hearing aid (HD) and the auxiliary device (AUX), indicated by transceiver unit WLR.sub.2 in the hearing aid. An interaural link may be established between two hearing aids of a binaural hearing system using respective wireless transceivers WLR.sub.1, e.g. to exchange audio data, to avail a multitude of microphone signals (or parts thereof) for an algorithm according to the present disclosure to precisely determine a direction (DOA) to a target sound source.

[0137] FIG. 5 shows a flow diagram for a method of operating a hearing device, e.g. a hearing aid according to the present disclosure.

[0138] The method provides a scheme for estimating the direction of arrival of sound emitted by one or more sound sources creating a sound field in the surroundings of the one or more sound sources. The method comprises: [0139] S1. providing an array of sound sensitive sensors, such as microphones, each sensor providing a stream of digital samples representing the sound field of the surroundings; [0140] S2. providing at least one sample of said surrounding sound field from each of said sensors at substantially the same time instant; [0141] S3. providing a parametric model configured to be able to describe the sound field as a function of the direction of arrival in a region surrounding and adjacent to said array; [0142] S4. optimizing said model with respect to its parameters based on said sound samples; and [0143] S5. minimizing a cost function of the model with respect to said direction of arrivals.

[0144] FIG. 6 shows an embodiment of a sound system (SS) for estimating the direction of arrival of a sound signal emitted from one or more sound sources according to the present disclosure. The sound system receives composite signals z.sub.1, . . . , z.sub.N at respective input transducers IT1, . . . , IT.sub.N from a surrounding sound field. The sound field comprises a time-varying mixture of one or more (target, e.g. speech) sound sources S.sub.j, j=1, N.sub.ss (e.g. one sound source, S, as indicated in FIG. 3A, 3B (i.e. N.sub.ss=1)) and additional noise, as indicated in FIG. 6 by input sound signal z(t)=s(t)+E(t), where E(t) represents noise. The sound system (SS) comprises a sound sensor unit (SSU) comprising an array of N sound receiving transducers (e.g. microphones), each providing an analogue electric input signal representative of the sound signal z.sub.1, z.sub.N received at a respective input transducer. The sound system (SS) further comprises a sampling unit (SU) for providing at least one sample of the surrounding sound field from each of electric input signals at substantially the same time instant. The sampling unit (SU) comprises analogue to digital converters AD.sub.1, . . . , AD.sub.N coupled to a respective one of the input transducers IT.sub.1, . . . , IT.sub.N, and providing respective digitized electric input signals as streams of samples z.sub.1, . . . , z.sub.N. The sound system (SS) further comprises a processing unit (PU) comprising a) a model unit (MU) comprising a parametric model configured to be able to describe the sound field at the array of N sound receiving transducers as a function of the direction of arrival (DOA) in a region surrounding and adjacent to the array, b) a model optimizing unit (MOU) configured to optimize said model with respect to its parameters based on said sound samples, c) a cost optimizing unit (COU) configured to minimize a cost function of the model with respect to said direction of arrivals (DOA); and an estimating unit (EU) configured to estimate the (final) direction of arrival (DOA) based on said parametric model with the optimized parameters and the optimized cost function. In an embodiment, the sampling unit is configured to sample the analogue electric input signals with a relatively low sampling frequency, e.g. below 1 kHz, such as below 500 Hz. Thereby a low pass filtering of the electric input signals is effectively provided. In an embodiment, the sampling frequency is configured to isolate a fundamental frequency of the electric input signals. In an embodiment, the sampling frequency is configured to isolate a fundamental frequency corresponding to a human voice from higher lying harmonics and providing respective filtered electric input signals z.sub.1, z.sub.N. In an embodiment, the sound system is configured to use such filtered electric input signals for identification of the direction of arrival.

[0145] FIG. 7 shows a simplified block diagram of an embodiment of a hearing device (HD) comprising a beamformer filtering unit (BFU) for providing a beamformed signal Y.sub.BF based on a multitude of electric sound signals z.sub.1, . . . , z.sub.N received from a sound system (SS) according to the present disclosure. Alternatively, the beamformer filtering unit may user electric input signals from other input transducers. The hearing device (HD) further comprises a signal processing unit (SPU) for processing the beamformed signal, e.g. according to a user's needs, e.g. to a user's hearing impairment and providing a processed signal ES, which is fed to an output unit (FBS, DA-OT) for providing a time-variant signal es(t) perceivable as sound to the user. FIG. 7 shows an embodiment of hearing device HD, e.g. a hearing aid, according to the present disclosure. A time variant input sound z(t) is assumed to comprise a mixture of a target signal component s(t) and a noise signal component (t) is picked up by the hearing device, processed and provided in a processed form to a user wearing the hearing device as an audible signal (cf. e.g. FIG. 3A, 3B). The sound source system (SS) of the hearing device of FIG. 7 comprises a sound sensor unit (SSU) comprising a multitude of input transducers IT.sub.j, j=1, . . . , N, each providing an (analogue) electric input signal representative of sound z.sub.i(t), and a sampling unit (SU) comprising a multitude of analogue to digital converters AD.sub.j providing digitized versions z.sub.1, . . . , z.sub.M of the electric input signal. The (digital) electric input signals z.sub.1, . . . , z.sub.M are fed to the processing unit (PU) of the sound source system (SS) and to respective analysis filter bank unit FBA.sub.j of the beamformer filtering unit (BFU). The processing unit (PU) processes the electric input signals z.sub.1, . . . , z.sub.M, according to a method of the present disclosure, and provides a (final) direction of arrival (DOA) that is fed to a control unit (CONT) of the beamformer filtering unit (BFU). The analysis filter bank units FBA.sub.j, j=1, . . . , N, are configured to convert the electric time-domain signal z.sub.j to a number of frequency sub-band signals (k=1, . . . , K), thereby providing the electric inputs signals z.sub.j, j=1, N, in a time-frequency representation Z.sub.j(k,m), k being a frequency sub-band index (k=1, . . . , K) and m being a time frame index. The multi-input beamformer filtering unit (BFU) further comprises a beamformer (BF) and a control unit CONT. The beamformer BF (and optionally the control unit CONT) receives the multitude of electric input signals Z.sub.i, i=1, . . . , N (or possibly other electric input signals representative of a current sound field around the user wearing the hearing device), and provides beamformed signal Y.sub.BF. The control unit CONT comprises a memory MEM wherein complex weights W.sub.ij can be stored. The complex weights W.sub.ij define possible pre-defined (or previously updated) fixed beam formers of the beamformer filtering unit BFU (fed to BF via signal W.sub.ij). The control unit (CONT) comprises a processor PRO, e.g. for updating look vector d and/or beamformer filtering weights W.sub.ij, e.g. influenced or based on the DOA. The control unit CONT further comprises one or more voice activity detectors VAD for estimating whether or not a given input signal (e.g. whether or not a given time-frequency unit of the input signal) comprises (or is dominated by) a voice. Respective control signals V-NV (voice-no voice) are used in the control unit CONT to determine the look vector d and/or beamformer filtering weights W.sub.ij, and is further fed to the beamformer BF. In an embodiment, the control unit CONT is configured to receive the multitude of electric input signals Z, i=1, N, from analysis filter banks FBA.sub.j and the beamformed signal Y.sub.BF from the beamformer BF. The signal Y.sub.BF comprises an estimate of the target signal component of the input sound field. The hearing device HD may further comprise a (single channel) post filtering unit receiving (spatially filtered) target signal estimate Y.sub.BF (and a corresponding spatially filtered) noise signal estimate, to provide a (further) noise reduced target signal estimate Y.sub.NR, e.g. based on knowledge of the noise part of the electric input signals Z.sub.i, i=1, . . . , N. The hearing aid further comprises a signal processing unit SPU for (further) processing the beamformed and possibly further noise reduced signal (possibly including the mentioned post filtering unit) and providing a processed signal ES. The signal processing unit SPU may be configured to apply a level and frequency dependent shaping of the noise reduced signal Y.sub.BF (Y.sub.NR), e.g. to compensate for a user's hearing impairment. The hearing aid further comprises a synthesis filter bank FBS for converting the processed frequency sub-band signal ES to a time domain signal es, which is fed to an output unit DA-OT (comprising digital to analogue converter DA and an output transducer OT) for providing stimuli es(t) to a user as a signal perceivable as sound. The output transducer OT may e.g. comprise a loudspeaker for presenting the processed signal es to the user as sound or a vibrator of a bone-conducting hearing device. The forward path from the input unit to the output unit of the hearing aid is here operated partly in the time-frequency domain (processed in a number of frequency sub-bands FB.sub.k, k=1, . . . , K). In another embodiment, the forward path from the input unit to the output unit of the hearing aid may be operated in the time domain. The hearing aid may further comprise a user interface and one or more detectors allowing user inputs and detector inputs to be received by the beamformer filtering unit, e.g. the control unit CONT (cf. e.g. FIG. 4). An adaptive functionality of the beamformer filtering unit BFU may be provided based on the current DOA estimated according to the present disclosure. In an embodiment, the beamformer filtering unit is configured to track a number of sound sources, by individually estimating a direction of arrival from each of the number of sound sources on a continuous basis.

[0146] In an embodiment (as shown in FIG. 7), at least some of (such as a majority or all of) the sound sensors (here microphones) used for the DOA estimation (in the sound system unit SS) are also used as inputs to the beamformer filtering unit. It should be noted that the (typical) minimum distance requirement of hearing aid microphones can be dispensed with when a multi-microphone array using the scheme of the present disclosure is used. A sound sensor array comprising a multitude N of microphones (e.g. N4 or N8), in any spatial configurationfixated in space relative to each otherwhen used to provide spatial domain filtering is not constrained to configurations of having the distance of a half-wavelength that is necessary in the delay-and-sum beamforming technology currently being used in hearing devices. According to the present disclosure, a flexible spatial configuration is possible.

[0147] With reference to FIG. 8 illustrating a table conversation situation with quasi constant location of a multitude of occasional talkers/listeners, look vectors (transfer functions from source to microphones) and/or filter weights corresponding to a particular DOA may be determined and stored in the memory unit MEM of the control unit CONT in FIG. 7, so that they can be quickly loaded into the beamformer (BF) when the talker of a given DOA is active (again). Thereby the non-active beams can be considered to represent virtual microphones that can be activated one at a time, or two or more simultaneously.

[0148] FIG. 8 shows an application of a hearing device or a hearing system according to the present disclosure for segregating individual sound sources in a multi-sound source environment. In FIG. 8, the sound sources are persons (that at a given time are talkers (S) or listeners (L)) located around a user (U, that at the time illustrated is a listener (L)). The user (U) may wear a sound system comprising a head-mounted array of microphones according to the present disclosure that allows segregation of each talker and allows the user to tune in depending on the person (S) that is currently speaking as indicated by (schematic) elliptic beams of angular width () sufficiently small to enclose one (and only one) of the persons surrounding the user. In the example of FIG. 8, the person speaking is indicated by S, and the sound system is focused on this person as indicated by direction of arrival (DOA) and a bold elliptic beam including the speaker (S). The angular width of the beam (main lobe) is preferably of the order of 30 or smaller, e.g. smaller than 25, cf. e.g. FIG. 9.

[0149] FIG. 9 schematically shows a plot of attenuation ([dB]) versus angle ([] (at a specific frequency, e.g. 2 kHz) for a sound system comprising an array of sound sensitive sensors according to the present disclosure. The plot illustrates a main lobe centered around a direction of arrival (DOA) and a number of side lobes. A (first) side lobe attenuation [dB] is indicated as the attenuation of a first side lobe compared to the main lobe. The attenuation of a first side lobe compared to the main lobe is e.g. arranged to be larger than 10 dB, such as larger than 15 dB. The angular width of the main lobe (denoted in FIG. 8) is preferably configurable, and e.g. configured to be smaller than 30, such as smaller than 20.

[0150] FIG. 10 illustrates a situation where the hearing device or hearing system (L-HD, R-HD) according to the present disclosure is configured to determine (and possibly track) a multitude of directions of arrival (here three, DOA1, DOA2, DOA3) to sound sources of interest (e.g. sources comprising modulated signals, e.g. speech, SL1, SL2, SL3) in the environment of the user (U) wearing the hearing device or hearing system. The directions of arrival (DOA1, DOA2, DOA3) as determined by the sound system as described above are used by a beamformer filtering unit of the hearing devices (L-HD, R-HD), individually and/or binaurally (cf. e.g. FIG. 7), to determine a beam pattern (BP) that attempts to maintain (target) sound from the (target) sound sources of interest (SL1, SL2, SL3) to the user (U) (and to attenuate sound from other directions). In an embodiment, the hearing device or hearing system is configured to track the active target sound sources over time and thus maintain focus of the beam pattern (BP) on the active sound sources (even if the user and sound sources move relative to each other). In an embodiment the tracking unit is configured to use near-field tracking as described in [Gustafsson et al., 2015]: In target tracking, a dynamic motion model of the form

x(t+1)=f(x(t),v(t))(20)

describes the motion of the target over time. The state x(t) includes the position (X(t), Y(t)) at time t, and v(t) denotes process noise. By letting the delay .sub.n(X(t), Y(t)) (see eq. (18) below) be time varying, a more or less standard nonlinear estimation problem is achieved, where the measurement equation is

z(t)=T.sup.T(X,Y)S(t)+(t)(19)

where T is the regression matrix (cf. eq. (10), (11) above), and where the delay is given by

[00013]$\begin{array}{cc}{}_n\left(X,Y\right)=\frac{1}{c}.Math.\sqrt{{\left(x_n-X\right)}^2+{\left(y_n-Y\right)}^2}& \left(18\right)\end{array}$

[0151] The present concepts can be used in many different applications, e.g. to control of the beam(s)/beam former(s), e.g. to steer the direction of an active beam, e.g. in combination with a direction selecting (pointer) device, e.g. using (possibly ear-based) electrooculography EarEOG, etc., cf. e.g. US20140098981A1 and [Manabe et al.; 2013]. EOG may be used to extract a user's (currently) preferred listening direction. Other means of assessing preference, e.g. a control wheel on a smart phone, may be used to define the preferred direction/DOA (). Using the concept of virtual microphones that can be activated at will, listening to a multitude of sound sources does necessarily require attenuation of all other sources. The preferred direction(s) can simply be emphasized, e.g. by 10 dB. Thereby the listening scene of the other sources is not lost.

[0152] Details of the DOA estimation method are provided in an article by the present inventors [Gustafsson et al., 2015], to which the equation numbers used in the present disclosure refer.

[0153] It should be appreciated that reference throughout this specification to one embodiment or an embodiment or an aspect or features included as may means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the disclosure. Furthermore, the particular features, structures or characteristics may be combined as suitable in one or more embodiments of the disclosure. The previous description is provided to enable any person skilled in the art to practice the various aspects described herein. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects. The claims are thus not intended to be limited to the aspects shown herein, but is to be accorded the full scope consistent with the language of the claims, wherein reference to an element in the singular is not intended to mean one and only one unless specifically so stated, but rather one or more. Unless specifically stated otherwise, the term some refers to one or more.

[0154] Accordingly, the scope should be judged in terms of the claims that follow.

REFERENCES

[0155] [Doron & Nevet; 2008]), Miriam A. Doron and Amir Nevet, Robust wavefield interpolation for adaptive wideband beamforming, Signal Processing, vol. 88, no. 6, pp. 1579-1594, June 2008. [0156] US20140098981A1 (OTICON) Oct. 4, 2014 [0157] [Manabe et al.; 2013] Hiroyuki Manabe, Masski Fukumoto, Tohru Yagi, Conductive Rubber Electrodes for Earphone-Based Eye Gesture Input Interface, ISWC'13, Sep. 9-12, 2013, Zurich, Switzerland, pp. 33-39. [0158] [Gustafsson et al., 2015] Fredrik Gustafsson, Gustaf Hendeby, David Lindgren, George Mathai, and Hans Habberstadt, Direction of Arrival Estimation in Sensor Arrays Using Local Series Expansion of the Received Signal, 18TH INTERNATIONAL CONFERENCE OF INFORMATION FUSION, 6 Jul. 2015, pages 761-766.