HEARING AID COMPRISING A BEAM FORMER FILTERING UNIT COMPRISING A SMOOTHING UNIT

Abstract

A hearing aid comprises a resulting beam former (Y) for providing a resulting beamformed signal Y.sub.BF based on first and second electric input signals IN.sub.1 and IN.sub.2, first and second sets of complex frequency dependent weighting parameters W.sub.11(k), W.sub.12(k) and W.sub.21(k), W.sub.22(k), and a resulting complex, frequency dependent adaptation parameter β(k)•β(k) may be determined as <C.sub.2*•C.sub.1>/<(|C2|.sup.2>+c), where * denotes the complex conjugation and • denotes the statistical expectation operator, and c is a constant, and wherein said adaptive beam former filtering unit (BFU) comprises a smoothing unit for implementing said statistical expectation operator by smoothing the complex expression C.sub.2*•C.sub.1 and the real expression |C.sub.2>.sup.2 over time. Alternatively, β(k) may be determined from the following expression

[00001]$\beta =\frac{w_{C.Math..Math.1}^H.Math.C_v.Math.w_{C.Math..Math.2}}{w_{C.Math..Math.2}^H.Math.C_v.Math.w_{C.Math..Math.2}},$

where w.sub.C1 and w.sub.C2 are the beamformer weights representing the first (C.sub.1) and the second (C.sub.2) beamformers, respectively, C.sub.v is a noise covariance matrix, and H denotes Hermitian transposition. Corresponding methods of operating a hearing aid, and a hearing aid utilizing smoothing β(k) based on adaptive covariance smoothing are disclosed.

Claims

1. A hearing aid adapted for being located in an operational position at or in or behind an ear or fully or partially implanted in the head of a user, the hearing aid comprising first and second microphones (M.sub.BTE1, M.sub.BTE2) for converting an input sound to first IN.sub.1 and second IN.sub.2 electric input signals, respectively, an adaptive beam former filtering unit (BFU) for providing a resulting beamformed signal Y.sub.BF, based on said first and second electric input signals, the adaptive beam former filtering unit comprising, a first memory comprising a first set of complex frequency dependent weighting parameters W.sub.11(k), W.sub.12(k) representing a first beam pattern (C1), where k is a frequency index, k=1, 2, . . . , K, a second memory comprising a second set of complex frequency dependent weighting parameters W.sub.21(k), W.sub.22(k) representing a second beam pattern (C2), where said first and second sets of weighting parameters W.sub.11(k), W.sub.12(k) and W.sub.21(k), W.sub.22(k), respectively, are predetermined and possibly updated during operation of the hearing aid, an adaptive beam former processing unit for providing an adaptively determined adaptation parameter β(k) representing an adaptive beam pattern (ABP) configured to attenuate unwanted noise as much as possible under the constraint that sound from a target direction is essentially unaltered, and a resulting beam former (Y) for providing said resulting beamformed signal Y.sub.BF based on said first and second electric input signals IN.sub.1 and IN.sub.2, said first and second sets of complex frequency dependent weighting parameters W.sub.11(k), W.sub.12(k) and W.sub.21(k), W.sub.22(k), and said resulting complex, frequency dependent adaptation parameter β(k), where β(k) may be determined as $\beta \left(k\right)=\frac{〈C_2^{*}.Math.C_1〉}{〈{.Math.C_2.Math.}^2〉+c^{\prime }}$ where * denotes the complex conjugation and • denotes the statistical expectation operator, and c is a constant, wherein said adaptive beam former filtering unit (BFU) comprises a smoothing unit for implementing said statistical expectation operator by smoothing the complex expression C.sub.2*•C.sub.1 and the real expression |C.sub.2|.sup.2 over time.

2. A hearing aid according to claim 1, wherein the smoothing unit is configured to apply substantially the same smoothing time constants for the smoothing of the complex expression C.sub.2*•C.sub.1 and the real expression |C.sub.2|.sup.2.

3. A hearing aid according to claim 1, wherein the smoothing unit is configured to smoothe a resulting adaptation parameter β(k).

4. A hearing aid according to claim 4, wherein the smoothing unit is configured to provide that the attack and release time constants involved in the smoothing of the resulting adaptation parameter β(k) is larger than the corresponding attack and release time constants involved in the smoothing of the complex expression C.sub.2*•C.sub.1 and the real expression |C.sub.2|.sup.2.

5. A hearing aid according to claim 1, wherein the smoothing unit is configured to provide that the attack and release time constants involved in the smoothing of the complex expression C.sub.2*•C.sub.1 and the real expression |C.sub.2|.sup.2 are adaptively determined.

6. A hearing aid according to claim 1, wherein the smoothing unit is configured to provide that the attack and release time constants involved in the smoothing of the resulting adaptation parameter β(k) are adaptively determined.

7. A hearing aid according to claim 1, wherein the smoothing unit comprises a low pass filter implemented as an IIR filter with a fixed time constant, and an IIR filter with a configurable time constant.

8. A hearing aid according to claim 7 wherein the smoothing unit is configured to determine the configurable time constant by a function unit providing a predefined function of the difference between a first filtered value of the real expression |C.sub.2|.sup.2 when filtered by an IIR filter with a first time constant, and a second filtered value of the real expression |C.sub.2|.sup.2 when filtered by an IIR filter with a second time constant, wherein the first time constant is smaller than the second time constant.

9. A hearing aid according to claim 8 wherein the function unit comprises an ABS unit providing an absolute value of the difference between the first and second filtered values.

10. A hearing aid according to claim 8 wherein the first and second time constants are fixed time constants.

11. A hearing aid according to claim 9 wherein the first time constant the fixed time constant and the second time constant is the configurable time constant.

12. A hearing aid according to claim 8 wherein the predefined function is a decreasing function of the difference between the first and second filtered values.

13. A hearing aid according to claim 12 wherein the predefined function is one of a binary function, a piecewise linear function, and a continuous monotonous function.

14. A hearing aid according to claim 8 wherein the smoothing unit comprises respective low pass filters implemented as IIR filters using said configurable time constant for filtering real and imaginary parts of the expression C.sub.2*•C.sub.1 and the real expression |C.sub.2|.sup.2, and wherein said configurable time constant is determined from |C.sub.2|.sup.2.

15. A hearing aid adapted for being located in an operational position at or in or behind an ear or fully or partially implanted in the head of a user, the hearing aid comprising first and second microphones (M.sub.BTE1, M.sub.BTE2) for converting an input sound to first IN.sub.1 and second IN.sub.2 electric input signals, respectively, an adaptive beam former filtering unit (BFU) for providing a resulting beamformed signal Y.sub.BF, based on said first and second electric input signals, the adaptive beam former filtering unit comprising, a first memory comprising a first set of complex frequency dependent weighting parameters W.sub.11(k), W.sub.12(k) representing a first beam pattern (C1), where k is a frequency index, k=1, 2, . . . , K, a second memory comprising a second set of complex frequency dependent weighting parameters W.sub.21(k), W.sub.22(k) representing a second beam pattern (C2), where said first and second sets of weighting parameters W.sub.11(k), W.sub.12(k) and W.sub.21(k), W.sub.22(k), respectively, are predetermined and possibly updated during operation of the hearing aid, an adaptive beam former processing unit for providing an adaptively determined adaptation parameter β(k) representing an adaptive beam pattern (ABP) configured to attenuate unwanted noise as much as possible under the constraint that sound from a target direction is essentially unaltered, and a resulting beam former (Y) for providing said resulting beamformed signal Y.sub.BF based on said first and second electric input signals IN.sub.1 and IN.sub.2, said first and second sets of complex frequency dependent weighting parameters W.sub.11(k), W.sub.12(k) and W.sub.21(k), W.sub.22(k), and said resulting complex, frequency dependent adaptation parameter β(k), wherein the adaptive beamformer processing unit is configured to determine the adaptation parameter β(k) from the following expression $\beta =\frac{w_{C.Math..Math.1}^H.Math.C_v.Math.w_{C.Math..Math.2}}{w_{C.Math..Math.2}^H.Math.C_v.Math.w_{C.Math..Math.2}},$ where w.sub.C1 and w.sub.C2 are the beamformer weights representing the first (C.sub.1) and the second (C.sub.2) beamformers, respectively, C.sub.v is the noise covariance matrix, and H denotes Hermitian transposition.

16. A hearing aid according to claim 1 comprising a hearing instrument adapted for being located at or in an ear of a user or for being fully or partially implanted in the head of a user, a headset, an earphone, an ear protection device or a combination thereof.

17. A method of operating a hearing aid adapted for being located in an operational position at or in or behind an ear or fully or partially implanted in the head of a user, the method comprising converting an input sound to, or providing, first IN.sub.1 and second IN.sub.2 electric input signals, adaptively providing a resulting beamformed signal Y.sub.BF, based on said first and second electric input signals; storing in a first memory a first set of complex frequency dependent weighting parameters W.sub.11(k), W.sub.12(k) representing a first beam pattern (C1), where k is a frequency index, k=1, 2, . . . , K; storing in a second memory comprising a second set of complex frequency dependent weighting parameters W.sub.21(k), W.sub.22(k) representing a second beam pattern (C2), wherein said first and second sets of weighting parameters W.sub.11(k), W.sub.12(k) and W.sub.21(k), W.sub.22(k), respectively, are predetermined and possibly updated during operation of the hearing aid, providing an adaptively determined adaptation parameter β(k) representing an adaptive beam pattern (ABP) configured to attenuate unwanted noise as much as possible under the constraint that sound from a target direction is essentially unaltered, and providing said resulting beamformed signal Y.sub.BF based on said first and second electric input signals IN.sub.1 and IN.sub.2, said first and second sets of complex frequency dependent weighting parameters W.sub.11(k), W.sub.12(k) and W.sub.21(k), W.sub.22(k), and said resulting complex, frequency dependent adaptation parameter β(k), where β(k) may be determined as $\beta \left(k\right)=\frac{〈C_2^{*}.Math.C_1〉}{〈{.Math.C_2.Math.}^2〉+c^{\prime }}$ where * denotes the complex conjugation and • denotes the statistical expectation operator, and c is a constant, and smoothing the complex expression C.sub.2*•C.sub.1 and the real expression |C.sub.2|.sup.2 over time.

18. A method of operating a hearing aid adapted for being located in an operational position at or in or behind an ear or fully or partially implanted in the head of a user, the method comprising converting an input sound to, or providing, first IN.sub.1 and second IN.sub.2 electric input signals, adaptively providing a resulting beamformed signal Y.sub.BF, based on said first and second electric input signals; storing in a first memory a first set of complex frequency dependent weighting parameters W.sub.11(k), W.sub.12(k) representing a first beam pattern (C1), where k is a frequency index, k=1, 2, . . . , K; storing in a second memory comprising a second set of complex frequency dependent weighting parameters W.sub.21(k), W.sub.22(k) representing a second beam pattern (C2), wherein said first and second sets of weighting parameters W.sub.11(k), W.sub.12(k) and W.sub.21(k), W.sub.22(k), respectively, are predetermined and possibly updated during operation of the hearing aid, providing an adaptively determined adaptation parameter β(k) representing an adaptive beam pattern (ABP) configured to attenuate unwanted noise as much as possible under the constraint that sound from a target direction is essentially unaltered, and providing said resulting beamformed signal Y.sub.BF based on said first and second electric input signals IN.sub.1 and IN.sub.2, said first and second sets of complex frequency dependent weighting parameters W.sub.11(k), W.sub.12(k) and W.sub.21(k), W.sub.22(k), and said resulting complex, frequency dependent adaptation parameter β(k), wherein said resulting complex, frequency dependent adaptation parameter β(k) is determined from the following expression $\beta =\frac{w_{C.Math..Math.1}^H.Math.C_v.Math.w_{C.Math..Math.2}}{w_{C.Math..Math.2}^H.Math.C_v.Math.w_{C.Math..Math.2}},$ where w.sub.C1 and w.sub.C2 are the beamformer weights representing the first (C.sub.1) and the second (C.sub.2) beamformers, respectively, C.sub.v is a noise covariance matrix, and H denotes Hermitian transposition.

19. A method according to claim 17 comprising adaptive smoothing of a covariance matrix for said electric input signals comprising adaptively changing time constants (τ.sub.att, τ.sub.rel) for said smoothing in dependence of changes (ΔC) over time in covariance of said first and second electric input signals; wherein said time constants have first values (τ.sub.att1, τ.sub.rel1) for changes in covariance below a first threshold value (ΔC.sub.th1) and second values (τ.sub.att2, τ.sub.rel2) for changes in covariance above a second threshold value (ΔC.sub.th2), wherein the first values are larger than corresponding second values of said time constants, while said first threshold value (ΔC.sub.th1) is smaller than or equal to said second threshold value (ΔC.sub.th2).

20. A method according to claim 18 comprising adaptively smoothing said noise covariance matrix C.sub.v comprising adaptively changing time constants (τ.sub.att, τ.sub.rel) for said smoothing in dependence of changes (ΔC) over time in covariance of said first and second electric input signals; wherein said time constants have first values (τ.sub.att1, τ.sub.rel1) for changes in covariance below a first threshold value (ΔC.sub.th1) and second values (τ.sub.att2, τ.sub.rel2) for changes in covariance above a second threshold value (ΔC.sub.th2), wherein the first values are larger than corresponding second values of said time constants, while said first threshold value (ΔC.sub.th1) is smaller than or equal to said second threshold value (ΔC.sub.th2).

21. A method according to claim 20 comprising that the noise covariance matrix is C.sub.v is updated when only noise is present.

22. Use of a hearing aid as claimed in claim 1.

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

Description

BRIEF DESCRIPTION OF DRAWINGS

[0122] 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:

[0123] FIG. 1 shows an adaptive beam former configuration, where the adaptive beam former in the k.sup.th frequency channel Y(k) is created by subtracting a target cancelling beam former scaled by the adaptation factor β(k) from an omnidirectional beam former,

[0124] FIG. 2 shows an adaptive beam former configuration similar to the one shown in FIG. 1, but where the adaptive beam pattern Y(k) is created by subtracting a target cancelling beam former C.sub.2(k) scaled by the adaptation factor β(k) from another fixed beampattern C.sub.1(k),

[0125] FIG. 3 shows an exemplary block diagram illustrating how the adaptation factor β is calculated from equation (1), which in the numerator contains the average value of C.sub.2*C.sub.1 and in the denominator contains the average value of C.sub.2*C.sub.2=|C.sub.2|.sup.2,

[0126] FIG. 4 shows a block diagram of a first order IIR filter, where the smoothing properties is controlled by a coefficient (coef),

[0127] FIG. 5A shows an example of smoothing of the input signal |C.sub.2|.sup.2, wherein a long time constant will provide a stable estimate, but the convergence time will be slow, if the level suddenly changes from a high level to a low level, and

[0128] FIG. 5B shows an example of smoothing of the input signal |C.sub.2|.sup.2, wherein the time constant is short, and have a fast convergence, when the level changes, but the overall estimate has higher variance,

[0129] FIG. 6 shows a block diagram illustrating how the low-pass filter given in FIG. 4 may be implemented with different attack and release coefficients,

[0130] FIG. 7 shows an exemplary block diagram illustrating how the adaptation factor β is calculated from equation (1), but compared to FIG. 3, we do not only low-pass filter C.sub.2*C.sub.1 and |C.sub.2|.sup.2, we also low-pass filter the calculated adaptation factor β,

[0131] FIG. 8A shows a first exemplary block diagram of an improved low-pass filter, and

[0132] FIG. 8B shows a second exemplary block diagram of an improved low-pass filter,

[0133] FIG. 9 shows the resulting estimate from the improved low-pass filter shown in FIG. 8A or 8B,

[0134] FIG. 10 shows an exemplary block diagram of an improved low-pass filter with a similar low-pass filter structure as in FIG. 8A, but in FIG. 10, the adaptive coefficient depends on the level changes of |C.sub.2 |.sup.2,

[0135] FIG. 11 shows an exemplary block diagram of an improved low-pass filter with a similar low-pass filter structure as in FIG. 10, but in the embodiment of FIG. 11 the adaptive coefficient (coef) is estimated from a difference between two low-pass filtered estimates of |C.sub.2|.sup.2 with fixed slow and fast time constants, respectively,

[0136] FIG. 12 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,

[0137] FIG. 13A shows a block diagram of a first embodiment of a hearing aid according to the present disclosure, and

[0138] FIG. 13B shows a block diagram of a second embodiment of a hearing aid according to the present disclosure,

[0139] FIG. 14 shows a flow diagram of a method of operating an adaptive beam former for providing a resulting beamformed signal Y.sub.BF of a hearing aid according to an embodiment of the present disclosure, and

[0140] FIGS. 15A, 15B and 15C illustrate a general embodiment of a variable time constant covariance estimator according to the present disclosure, wherein

[0141] FIG. 15A schematically shows a covariance smoothing unit according to the present disclosure comprising a pre-smoothing unit (PreS) and a variable smoothing unit (VarS).

[0142] FIG. 15B shows an embodiment of the pre-smoothing unit, and

[0143] FIG. 15C shows an embodiment of the variable smoothing unit (VarS) providing adaptively smoothed of covariance estimators C.sub.x11 (m), C.sub.x12 (m), and C.sub.x22(m) according to the present disclosure.

[0144] FIGS. 16A, 16B, 16C and 16D illustrate a general embodiment of a variable time constant covariance estimator according to the present disclosure, wherein

[0145] FIG. 16A schematically shows a covariance smoothing unit according to the present disclosure based on beamformed signals C1, C2.

[0146] FIG. 16B shows an embodiment of the pre-smoothing unit based on beamformed signals C1, C2,

[0147] FIG. 16C shows an embodiment of the variable smoothing unit (VarS) adapted to the pres-smoothing unit of FIG. 16B, and

[0148] FIG. 16D schematically illustrates the determination of β based on smoothed covariance matrices (<|C2|.sup.2>, <C1C2*>) according to the present disclosure;

[0149] FIG. 17A schematically illustrates a first embodiment of the determination of β based on smoothed covariance matrices according to the present disclosure (compare FIG. 3), and

[0150] FIG. 17B schematically illustrates a second embodiment of the determination of β based on smoothed covariance matrices and further smoothing according to the present disclosure (compare FIG. 7), and

[0151] FIG. 18 schematically illustrates a third embodiment of the determination of β according to the present disclosure.

[0152] The figures are schematic and simplified for clarity, and they just show details which are essential to the understanding of the disclosure, while other details are left out. Throughout, the same reference signs are used for identical or corresponding parts.

[0153] Further scope of applicability of the present disclosure will become apparent from the detailed description given hereinafter. However, it should be understood that the detailed description and specific examples, while indicating preferred embodiments of the disclosure, are given by way of illustration only. Other embodiments may become apparent to those skilled in the art from the following detailed description.

DETAILED DESCRIPTION OF EMBODIMENTS

[0154] The detailed description set forth below in connection with the appended drawings is intended as a description of various configurations. The detailed description includes specific details for the purpose of providing a thorough understanding of various concepts. However, it will be apparent to those skilled in the art that these concepts may be practised without these specific details. Several aspects of the apparatus and methods are described by various blocks, functional units, modules, components, circuits, steps, processes, algorithms, etc. (collectively referred to as “elements”). Depending upon particular application, design constraints or other reasons, these elements may be implemented using electronic hardware, computer program, or any combination thereof.

[0155] The electronic hardware may include microprocessors, microcontrollers, digital signal processors (DSPs), field programmable gate arrays (FPGAs), programmable logic devices (PLDs), gated logic, discrete hardware circuits, and other suitable hardware configured to perform the various functionality described throughout this disclosure. Computer program shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executables, threads of execution, procedures, functions, etc., whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.

[0156] The present application relates to the field of hearing aids, e.g. hearing aids. FIGS. 1 and 2 shows respective two-microphone beam former configurations for providing a spatially filtered (beamformed) signal Y(k) in a number K of frequency sub-bands k=1, 2, . . . , K. The frequency sub-band signals X.sub.1(k), X.sub.2(k) are provided by analysis filter banks (Filterbank) base don the respective (digitized) microphone signals. The two beam formers C.sub.1(k) and C.sub.2(k) are provided by respective combination units (multiplication units ‘×’ and summation unit ‘+’) as (complex) linear combinations of the input signals:

C.sub.1(k)=w.sub.11(k)•X.sub.1(k)+w.sub.12(k)•X.sub.2(k)

C.sub.2(k)=w.sub.21(k)•X.sub.1(k)+w.sub.22(k)•X.sub.2(k)

[0157] FIG. 1 shows an adaptive beam former configuration, where the adaptive beam former in the k.sup.th frequency channel Y(k) is created by subtracting a target cancelling beam former C.sub.2(k) scaled by the adaptation factor β(k) from an omnidirectional beam former C.sub.1(k). In other words, Y(k)=C.sub.1(k)−(β•C.sub.2(k). The two beam formers C.sub.1, C.sub.2 are preferably orthogonal in the sense that [w.sub.11w.sub.12][w.sub.21w.sub.22].sup.H=0.

[0158] FIG. 2 shows an adaptive beam former configuration similar to the one shown in FIG. 1, but where the adaptive beam pattern Y(k) is created by subtracting a target cancelling beam former C.sub.2(k) scaled by the adaptation factor β(k) from another fixed beampattern C.sub.1(k). Whereas the C.sub.1(k) in FIG. 1 is an omnidirectional beampattern, the beampattern here is a beam former with a null towards the opposite direction of C.sub.2(k) as indicated in FIG. 2 by cardioid symbols adjacent to the C.sub.1(k) and C.sub.2(k) references. Other sets of fixed beampatterns C.sub.1(k) and C.sub.2(k) may as well be used.

[0159] An adaptive beampattern (Y(k)), for a given frequency band k, is obtained by linearly combining two beam formers C.sub.1(k) and C.sub.2(k). C.sub.1(k) and C.sub.2(k) are different (possibly fixed) linear combinations of the microphone signals.

[0160] The beampatterns could e.g. be the combination of an omnidirectional delay-and-sum-beam former C.sub.1(k) and a delay-and-subtract-beam former C.sub.2(k) with its null direction pointing towards the target direction (target cancelling beam former) as shown in FIG. 1 or it could be two delay-and-subtract-beam formers as shown in FIG. 2, where the one C.sub.1(k) has maximum gain towards the target direction, and the other beam former is a target cancelling beam former. Other combinations of beam formers may as well be applied. Preferably, the beam formers should be orthogonal, i.e. [w.sub.11w.sub.12][w.sub.21w.sub.22].sup.H=0. The adaptive beampattern arises by scaling the target cancelling beam former C.sub.2(k) by a complex-valued, frequency-dependent, adaptive scaling factor β(k) and subtracting it from the C.sub.1(k), i.e.

Y(k)=C.sub.1(k)−β(k)C.sub.2(k).

[0161] The beam former is adapted to work optimally in situations where the microphone signals consist of a point-noise target sound source in the presence of additive noise sources. Given this situation, the scaling factor β(k) is adapted to minimize the noise under the constraint that the sound impinging from the target direction is unchanged. For each frequency band k, the adaptation factor β(k) can be found in different ways. The solution may be found in closed form as

[00009]$\begin{array}{cc}\beta \left(k\right)=\frac{〈C_2^{*}.Math.C_1〉}{〈{.Math.C_2.Math.}^2〉+c},& \left(1\right)\end{array}$

where * denote the complex conjugation and • denotes the statistical expectation operator, which may be approximated in an implementation as a time average. As an alternative, the adaptation factor may be updated by an LMS or NLMS equation:

[00010]$\beta \left(n,k\right)=\beta \left(n-1,k\right)+\mu .Math.\frac{C_2^{*}.Math.Y-\alpha .Math..Math.\beta \left(n-1,k\right)}{{.Math.C_2.Math.}^2},$

[0162] In the following we omit the frequency channel index k. In (1), the adaptation factor β is estimated by averaging across the input data. A simple way to average across data is by low-pass filtering the data as shown in FIG. 3.

[0163] FIG. 3 shows a block diagram illustrating how the adaptation factor β is calculated from equation (1), which in the numerator contains the average value of C.sub.2*C.sub.1 and in the denominator contains the average value of C.sub.2*C.sub.2=|C.sub.2|.sup.2. We obtain the average value by low-pass filtering the two term. As C.sub.2*C.sub.1 typically is complex-numbered, we low-pass filter the real and the imaginary part of C.sub.2*C.sub.1 separately. In an embodiment, we low-pass filter the magnitude and the phase of C.sub.2*C.sub.1 separately. The resulting adaptation factor β is determined from input beam former signals C.sub.1 and C.sub.2 by appropriate functional units implementing the algebraic functions of equation (1), i.e. complex conjugation unit conj providing C.sub.2* from input C.sub.2, multiplication unit (‘×’) providing complex product C.sub.1.C.sub.2* from inputs C.sub.1 and C.sub.2*. Magnitude squared unit |•|.sup.2 provides magnitude squared |C.sub.2|.sup.2 of input C.sub.2. Complex and real valued sub-band signals C.sub.1•C.sub.2* and |C.sub.2|.sup.2, respectively, are low pass filtered by low pass filtering units LP to provide the resulting numerator and denominator in the expression for β in equation (1) (the constant c being added to the real value of |C.sub.2|.sup.2 by summation unit ‘+’ before or after the LP-filter (here after) to provide the expression for the denominator. The resulting adaptation factor β is provided by division unit ‘•/•’ based on inputs num (numerator) and den (denominator).

[0164] Such a low-pass filter LP may e.g. be implemented by a first order IIR filter as shown in FIG. 4. The IIR filter is implemented by summation units‘+’ delay element z.sup.−1 and multiplication unit ‘×’ for introducing a (possibly variable) smoothing element. FIG. 4 shows a first order IIR filter, where the smoothing properties is controlled by a coefficient (coef). The coefficient may take values between 0 and 1. A coefficient close to 0 applies averaging with a long time constant while a coefficient close to 1 applies a short time constant. In other words, if the coefficient is close to 1, only a small amount of smoothing is applied, while a coefficient close to 0 applies a higher amount of smoothing to the input signal. Averaging by a first order IIR filter have an exponential decay. As we apply smoothing on the inputs (|C.sub.2|.sup.2 and the real and imaginary part of C.sub.2*C.sub.1), the convergence of the adaptation factor β will be slow if the input level suddenly changes from a high level to a low level.

[0165] This is illustrated in FIGS. 5A and 5B showing a level (level) change from higher to lower and a corresponding time dependence (time) of a smoothed estimate depending on the smoothing coefficients of the LP-filter. FIG. 5A shows an example of smoothing of the input signal |C.sub.2|.sup.2, wherein a long time constant will provide a stable estimate, but the convergence time will be slow, if the level suddenly changes from a high level to a low level. By choosing a smaller time constant, a faster convergence can be achieved, but it the estimate will also have a higher variance. This is illustrated in FIG. 5B, which shows an example of smoothing of the input signal |C.sub.2|.sup.2, wherein the time constant is short, providing a fast convergence, when the level changes, but the overall estimate has higher variance.

[0166] We propose different ways to overcome this problem. A simple extension is to enable different attack and release coefficients in the low-pass filter. Such a low-pass filter is shown in FIG. 6.

[0167] FIG. 6 shows a block diagram illustrating how the low-pass filter given in FIG. 4 may be implemented with different attack and release coefficients. The different time constants are applied depending on whether the input is increasing (attack) or decreasing (release). Hereby it is possible to adapt fast in case of a sudden level change. Different attack and release times will however result in a biased estimate.

[0168] FIG. 7 shows an exemplary block diagram illustrating how the adaptation factor β is calculated from equation (1), but compared to FIG. 3, we do not only low-pass filter C.sub.2*C.sub.1 and |C.sub.2|.sup.2, we also low-pass filter the calculated adaptation factor β. It has the advantage that while the average value of C.sub.2*C.sub.1 and |C.sub.2|.sup.2 is sensitive to level drops, the low-pass filtering of β is not. We may thus move parts of the smoothing from C.sub.2*C.sub.1 and |C.sub.2|.sup.2 to β. Hereby we may allow more variance on the C.sub.2*C.sub.1 and |C.sub.2|.sup.2 estimates by applying smaller time constants. We thus obtain a faster convergence in the case, where the input level suddenly decreases. In FIG. 7, we propose not only smoothing the numerator and denominator for the β-estimation. We also smooth the estimated value of (β, i.e.

[00011]$\beta \left(k\right)=\frac{〈C_2^{*}.Math.C_1〉}{〈{.Math.C_2.Math.}^2〉+c},$

[0169] The advantage of smoothing the estimate of β is that the estimate is less sensitive to sudden drops in input level. Consequently, we can apply a shorter time constant to the low-pass filters used in the numerator and the denominator of (1). Hereby we can adapt faster in case of a sudden decreasing level. By post-smoothing β, we cope with the increased estimation variance.

[0170] Another option is to apply an adaptive smoothing coefficient that changes if a sudden input level change is detected. Embodiments of such low-pass filters are shown in FIGS. 8A and 8B.

[0171] FIG. 8A shows a first exemplary block diagram of an improved low-pass filter. The low-pass filter is able to change its time constant (or the equivalent coefficient (cod)) based on the difference between the input signal (Input) filtered by a low-pass filter (IIR-filter, cf. FIG. 4) having a (e.g. fixed) fast time constant and the input signal filtered by a low-pass filter having a (variable) slower time constant. If the difference Alnput between the two low-pass filters is high, it indicates a sudden change of the input level. This change of input level will enable a change of the time constant of the low-pass filter with the slow time constant to a faster time constant (the mapping function shown in the function block (fcn) indicating a change from slow to fast adaptation (larger to smaller time constants) with increasing input signal difference ΔInput. Hereby the low-pass filter will be able to adapt faster when we see sudden input level changes happen. If we only see small changes to the input level, a slower time constant is applied. By filtering the input signal by low-pass filters having different time constants (cf. LP-filtered Input) we will be able to detect when the level suddenly changes. Based on the level difference, we may adjust the coefficient by a non-linear function (fen in FIG. 8A). In an embodiment the non-linear function changes between a slow and a fast time constant, if the absolute difference between the signals are greater than a given threshold. Whenever a sudden level change is detected, the smoothing coefficient changes from a slow time constant to a faster time constant, hereby allowing a fast convergence until the new input level is reached. When the estimate has converged, the time constant returns to its slower value. Hereby we obtain not only a fast convergence but also less variance on the estimate when the input level does not fluctuate. To allow the function unit to work on positive as well as negative level changes (a well as directly on a complex signal) the function unit comprises a magnitude unit |•| that precedes the ΔInput to time constant mapping function.

[0172] FIG. 8B shows a second exemplary block diagram of an improved low-pass filter. The embodiment is similar to the embodiment of FIG. 8A, but the input difference signal is generated on the basis of two filtered signals with fixed fast and slow smoothing coefficients, and the resulting adapted smoothing coefficient (coef) is used to control the smoothing of a separate IIR filter that provides the LP-filtered input.

[0173] The resulting smoothing estimate from the low-pass filter shown in FIG. 8A or 8B is shown in FIG. 9. When an input level change is detected, the time constant is adapted to change from slow adaptation to a faster convergence (compared to the dashed line showing the slower convergence, cf. FIG. 5A). As soon as the estimate has adapted to the new level, the time constant is changed back to the slower value. Hereby we obtain faster convergence (compared to the dashed line showing the convergence using the slower time constant).

[0174] FIG. 10 shows an exemplary block diagram of an improved low-pass filter with a similar low-pass filter structure as in FIG. 8A, but in FIG. 10, the adaptive coefficient depends on the level changes of |C.sub.2|.sup.2. When low-pass filtering the numerator and the denominator of Equation (1), it is important that the same time constant is applied in both the numerator and the denominator. Here we propose that the adaptive coefficient depends on the level changes of |C.sub.2|.sup.2. In FIG. 10, the adaptive time constant is used as coefficient for the slow low-pass filter.

[0175] FIG. 11 shows an exemplary block diagram of an improved low-pass filter with a similar low-pass filter structure as in FIG. 10, but in the embodiment of FIG. 11 the adaptive coefficient (coef) is estimated from a difference between two low-pass filtered estimates of |C.sub.2|.sup.2 with fixed slow and fast time constants, respectively (cf. FIG. 8B). In FIG. 11, separate low-pass filters with fixed fast and fixed slow time constants are used to estimate the adaptive coefficient. Also other factors may be used to control the coefficient of the low-pass filters. E.g. a voice activity detector may be used to halt the update (by setting the coefficient to 0). In that case, the adaptive coefficient is solely updated during speech pauses.

[0176] FIG. 12 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.

[0177] FIG. 12 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 (e.g. exemplifying a hearing aid (HD) as shown in FIG. 13A, 13B). The BTE-part (BTE) and the ITE-part (ITE) are connected (e.g. electrically connected) by a connecting element (IC). In the embodiment of a hearing aid of FIG. 12, the BTE part (BTE) comprises two input transducers (here microphones) (M.sub.BTE1, M.sub.BTE2) 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. 12, 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. The hearing aid of FIG. 12 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 beam former 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 (cf. signal OUT in FIG. 13A, 13B), which is intended to be presented to a user. In the embodiment of a hearing aid device in FIG. 12, 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 (M.sub.BTE1, M.sub.BTE2). In another embodiment, the hearing aid may comprise only the ITE-microphone (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.

[0178] The hearing aid (HD) exemplified in FIG. 12 is a portable device and further comprises a battery (BAT) for energizing electronic components of the BTE- and ITE-parts.

[0179] The hearing aid (HD) comprises a directional microphone system (beam former 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. In an embodiment, the beam former filtering unit is adapted to receive inputs from a user interface (e.g. a remote control or a smartphone) regarding the present target direction. The memory unit (MEM) may e.g. comprise predefined (or adaptively determined) complex, frequency dependent constants (W.sub.ij) defining predefined or (or adaptively determined) ‘fixed’ beam patterns (e.g. omni-directional, target cancelling, etc.), together defining the beamformed signal Y.sub.BF (cf. e.g. FIG. 13A, 13B).

[0180] The hearing aid of FIG. 12 may constitute or form part of a hearing aid and/or a binaural hearing aid system according to the present disclosure.

[0181] The hearing aid (HD) according to the present disclosure may comprise a user interface UI, e.g. as shown in FIG. 12 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. 12, the screen of the user interface (UI) illustrates a Smooth beamforming APP. Parameters that govern or influence the current smoothing of adaptive beamforming, here fast and slow smoothing coefficients of low pass filters involved in the determination of the adaptive beamformer parameter β (cf. discussion in connection with FIGS. 8A, 8B, and FIG. 10, 11) can be controlled via the Smooth beamforming APP (with the subtitle: ‘Directionality. Configure smoothing parameters’). The smoothing parameters ‘Fast coefficient’ and ‘Slow coefficient’ can be set via respective sliders to a value between a minimum value (0) and a maximum value (1). The currently set values (here 0.8 and 0.2, respectively) are shown on the screen at the location of the slider on the (grey shaded) bar that span the configurable range of values. The coefficients could as well be shown as derived parameters such as time constants or other descriptions such as “calm” or “aggressive”. The coefficient can be derived from the time constant as coef =1-exp(−1/(f.sub.s *r)), where f.sub.s is the sample rate of the time frame, and τ is a time constant. The arrows at the bottom of the screen allow changes to a preceding and a proceeding screen of the APP, and a tab on the circular dot between the two arrows brings up a menu that allows the selection of other APPs or features of the device.

[0182] The auxiliary device and the hearing aid are adapted to allow communication of data representative of the currently selected direction (if deviating from a predetermined direction (already stored in the hearing aid)) to the hearing aid via a, e.g. wireless, communication link (cf. dashed arrow WL2 in FIG. 12). 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.

[0183] FIG. 13A shows a block diagram of a first embodiment of a hearing aid according to the present disclosure. The hearing aid of FIG. 13A may e.g. comprise a 2-microphone beam former configuration as e.g. shown in FIG. 1, 2, and a signal processing unit (SPU) for (further) processing the beamformed signal Y.sub.BF and providing a processed signal OUT. The signal processing unit may be configured to apply a level and frequency dependent shaping of the beamformed signal, e.g. to compensate for a user's hearing impairment. The processed signal (OUT) is fed to an output unit for presentation to a user as a signal perceivable as sound. In the embodiment of FIG. 13A, the output unit comprises a loudspeaker (SPK) for presenting the processed signal (OUT) to the user as sound. The forward path from the microphones to the loudspeaker of the hearing aid may be operated in the time domain. The hearing aid may further comprise a user interface (UI) and one or more detectors (DET) allowing user inputs and detector inputs (e.g. from a user interface as illustrated in FIG. 12) to be received by the beam former filtering unit (BFU). Thereby an adaptive functionality of the resulting adaptation parameter β may be provided.

[0184] FIG. 13B shows a block diagram of a second embodiment of a hearing aid according to the present disclosure. The hearing aid of FIG. 13B is similar in functionality to the hearing aid of FIG. 13A, also comprising a 2-microphone beam former configuration as e.g. shown in FIG. 1, 2, but the signal (where time-domain input signals IN.sub.1 and IN.sub.2 are provided as frequency sub-band signals IN.sub.1(k) and IN.sub.2(k), respectively, where k=1, 2, . . . , K, by respective analysis filter banks FBA1 and FBA2. Hence, the processing unit (SPU) for (further) processing the beamformed signal Y.sub.BF(k) is configured to process the beamformed signal Y.sub.BF(k) in a number (K) of frequency bands and providing processed (sub-band) signals OU(k), k=1, 2, . . . , K. The signal processing unit may be configured to apply a level and frequency dependent shaping of the beamformed signal, e.g. to compensate for a user's hearing impairment (and/or a challenging acoustic environment). The processed frequency band signals OU(k) are fed to a synthesis filter bank FBS for converting the frequency band signals OU(k) to a single time-domain processed (output) signal OUT, which is fed to an output unit for presentation to a user as a stimulus perceivable as sound. In the embodiment of FIG. 13B, the output unit comprises a loudspeaker (SPK) for presenting the processed signal (OUT) to the user as sound. The forward path from the microphones (M.sub.BTE1, M.sub.BTE2) to the loudspeaker (SPK) of the hearing aid is (mainly) operated in the time-frequency domain (in K frequency sub-bands).

[0185] FIG. 14 shows a flow diagram of a method of operating an adaptive beam former for providing a resulting beamformed signal Y.sub.BF of a hearing aid according to an embodiment of the present disclosure.

[0186] The method is configured to operate a hearing aid adapted for being located in an operational position at or in or behind an ear or fully or partially implanted in the head of a user.

[0187] The method comprises

[0188] S1. converting an input sound to first IN.sub.1 and second IN.sub.2 electric input signals,

[0189] S2. adaptively providing a resulting beamformed signal Y.sub.BF, based on said first and second electric input signals;

[0190] S3. storing in a first memory a first set of complex frequency dependent weighting parameters W.sub.11(k), W.sub.12(k) representing a first beam pattern (C1), where k is a frequency index, k=1, 2, . . . , K; storing in a second memory comprising a second set of complex frequency dependent weighting parameters W.sub.21(k), W.sub.22(k) representing a second beam pattern (C2), wherein said first and second sets of weighting parameters W.sub.11(.sub.k), W.sub.12(k) and W.sub.21(k), W.sub.22(k), respectively, are predetermined and possibly updated during operation of the hearing aid,

[0191] S4. providing an adaptively determined adaptation parameter β(k) representing an adaptive beam pattern (ABP) configured to attenuate unwanted noise as much as possible under the constraint that sound from a target direction is essentially unaltered, and

[0192] S5. providing said resulting beamformed signal Y.sub.BF based on said first and second electric input signals IN.sub.1 and IN.sub.2, said first and second sets of complex frequency dependent weighting parameters W.sub.11(k), W.sub.12(k) and W.sub.21(k), W.sub.22(k), and said resulting complex, frequency dependent adaptation parameter β(k), where (β(k) may be determined as

[00012]$\beta \left(k\right)=\frac{〈C_2^{*}.Math.C_1〉}{〈{.Math.C_2.Math.}^2〉+c^{\prime }}$

where * denotes the complex conjugation and • denotes the statistical expectation operator, and c is a constant

[0193] S6. smoothing the complex expression C.sub.2*•C.sub.1 and the real expression |C.sub.2|.sup.2 over time.

[0194] A Method of Adaptive Covariance Matrix Smoothing for Accurate Target Estimation and Tracking.

[0195] In a further aspect of the present disclosure, a method of adaptively smoothing covariance matrices is outlined in the following. A particular use of the scheme is for (adaptively) estimating a direction of arrival of sound from a target sound source to a person (e.g. a user of a hearing aid, e.g. a hearing aid according to the present disclosure).

[0196] The method is exemplified as an alternative scheme for smoothing of the adaptation parameter β(k) according to the present disclosure (cf. FIG. 16A-16D and 17A, 17B).

[0197] Signal Model:

[0198] We consider the following signal model of the signal x impinging on the i.sup.th microphone of a microphone array consisting of M microphones:

x.sub.i(n)=s.sub.i(n)+v.sub.i(n),   (1)

where s is the target signal, v is the noise signal, and n denotes the time sample index. The corresponding vector notation is

x(n)=s(n)+v(n),   (2)

where x(n)=[x.sub.1(n); x.sub.2(n), . . . , x.sub.M(n)].sup.T. In the following, we consider the signal model in the time frequency domain. The corresponding model is thus given by

X(k,m)=S(k,m)+V(k,m),   (3)

where k denotes the frequency channel index and m denotes the time frame index. Likewise X(k,m)=[X.sub.1(k,m), X.sub.2(k,m), . . . , X.sub.M(k,m)].sup.T. The signal at the i.sup.th microphone, x.sub.i is a linear mixture of the target signal s.sub.i and the noise v.sub.i. v.sub.i is the sum of all noise contributions from different directions as well as microphone noise. The target signal at the reference microphone s.sub.ref is given by the target signal s convolved by the acoustic transfer function h between the target location and the location of the reference microphone. The target signal at the other microphones is thus given by the target signal at the reference microphone convolved by the relative transfer function d=[1,d.sub.2, . . . , d.sub.M].sup.T between the microphones, i.e. s.sub.i=s*h*d.sub.i. The relative transfer function d depends on the location of the target signal. As this is typically the direction of interest, we term d the look vector. At each frequency channel, we thus define a target power spectral density (k, m) at the reference microphone, i.e.

σ.sub.s.sup.2(k, m)=|S(k, m)H(k, m)|.sup.2=|S(k, m).sub.ref|.sup.2,   (4)

where •denotes the expected value. Likewise, the noise spectral power density at the reference microphone is given by

σ.sub.v.sup.2(k, m)=|V(k, m).sub.ref|.sup.2,   (5)

[0199] The inter-microphone cross-spectral covariance matrix at the k.sup.th frequency channel for the clean signal s is then given by

C.sub.s(k, m)=σ.sub.s(k, m)d(k, m)d.sup.H(k, m),   (6)

where H denotes Hermitian transposition. We notice the M×M matrix C.sub.s(k,m) is a rank 1 matrix, as each column of C.sub.s(k,m) is proportional to d(k,m). Similarly, the inter-microphone cross-power spectral density matrix of the noise signal impinging on the microphone array is given by,

C.sub.v(k, m)=σ(k, m)Γ(k, m.sub.0), m>m.sub.0   (7)

where Γ(k, m.sub.0) is the M×M noise covariance matrix of the noise, measured some time in the past (frame index m.sub.0). Since all operations are identical for each frequency channel index, we skip the frequency index k for notational convenience wherever possible in the following. Likewise, we skip the time frame index m, when possible. The inter-microphone cross-power spectral density matrix of the noisy signal is then given by

C=C.sub.s+C.sub.v   (8)

C=σ.sub.s.sup.2dd.sup.11+σ.sub.v.sup.2Γ  (9)

where the target and noise signals are assumed to be uncorrelated. The fact that the first term describing the target signal, C is a rank-one matrix implies that the beneficial part (i.e., the target part) of the speech signal is assumed to be coherent/directional. Parts of the speech signal, which are not beneficial, (e.g., signal components due to late-reverberation, which are typically incoherent, i.e., arrive from many simultaneous directions) are captured by the second term.

[0200] Covariance Matrix Estimation

[0201] A look vector estimate can be found efficiently in the case of only two microphones based on estimates of the noisy input covariance matrix and the noise only covariance matrix. We select the first microphone as our reference microphone. Our noisy covariance matrix estimate is given by

[00013]$\begin{array}{cc}\hat{C}=\left[\begin{array}{cc}{\hat{C}}_{x.Math..Math.11}& {\hat{C}}_{x.Math..Math.12}\\ {\hat{C}}_{x.Math..Math.12}^{*}& {\hat{C}}_{x.Math..Math.22}\end{array}\right]& \left(10\right)\end{array}$

where * denotes complex conjugate. Each element of our noisy covariance matrix is estimated by low-pass filtering the outer product of the input signal, XX.sup.H. We estimate each element by a first order IIR low-pass filter with the smoothing factor α∈[0; 1], i.e.

[00014]$\begin{array}{cc}{\hat{C}}_x\left(m\right)=\{\begin{array}{c}\alpha .Math.{\hat{C}}_x\left(m-1\right)+\left(1-\alpha \right).Math.X\left(m\right).Math.{X\left(m\right)}^H,\mathrm{Target}.Math..Math.\mathrm{present}\\ \gamma .Math.{\hat{C}}_x\left(m-1\right)+\left(1-\gamma \right).Math.{\hat{C}}_{\mathrm{no}},\mathrm{Otherwise}\end{array};& \left(11\right)\end{array}$

[0202] We thus need to low-pass filter four different values (two real and one complex value), i.e. Ĉ.sub.x11(m), Re{Ĉ.sub.x12(m)}, Im{Ĉ.sub.x12(m)}, and Ĉ.sub.x22(m). We don't need Ĉ.sub.x21(m) since Ĉ.sub.x 21(m)=Ĉ.sub.12*. It is assumed that the target location does not change dramatically in speech pauses, i.e. it is beneficial to keep target information from previous speech periods using a slow time constant giving accurate estimates. This means that Ĉ.sub.x is not always updated with the same time constant and does not converge to Ĉ.sub.v in speech pauses, which is normally the case. In long periods with speech absence, the estimate will (very slowly) converge towards to C.sub.no using a smoothing factor close to one. The covariance matrix C could represent a situation where the target DOA is zero degrees (front direction), such that the system prioritizes the front direction when speech is absent. C, may e.g. be selected as an initial value of C.

[0203] In a similar way, we estimate the elements in the noise covariance matrix, in that case

[00015]$\begin{array}{cc}{\hat{C}}_v\left(m\right)=\{\begin{array}{c}{\alpha }_v.Math.{\hat{C}}_v\left(m-1\right)+\left(1-{\alpha }_v\right).Math.X\left(m\right).Math.{X\left(m\right)}^H,\mathrm{Noise}.Math..Math.\mathrm{only}\\ {\hat{C}}_v\left(m-1\right),\mathrm{Otherwise}\end{array};& \left(12\right)\end{array}$

[0204] The noise covariance matrix is updated when only noise is present. Whether the target is present or not may be determined by a modulation-based voice activity detector. It should be noted that “Target present” (cf. FIG. 15C) is not necessarily the same as the inverse of “Noise Only”. The VAD indicators controlling the update could be derived from different thresholds on momentary SNR or Modulation Index estimates.

[0205] Adaptive Smoothing

[0206] The performance of look vector estimation is highly dependent on the choice of smoothing factor α, which controls the update rate of Ĉ.sub.x(m). When a is close to zero, an accurate estimate can be obtained in spatially stationary situations. When α is close to 1, estimators will be able to track fast spatial changes, for example when tracking two talkers in a dialogue situation. Ideally, we would like to obtain accurate estimates and fast tracking capabilities which is a contradiction in terms of the smoothing factor and there is a need to find a good balance. In order to simultaneously obtain accurate estimates in spatially stationary situations and fast tracking capabilities, an adaptive smoothing scheme is proposed.

[0207] In order to control a variable smoothing factor, the normalized covariance

ρ(m)=C.sub.x11.sup.−1C.sub.x12,   (13)

can be observed an indicator for changes in the target DOA (where C.sub.x11.sup.−1 and C.sub.x12 are complex numbers).

[0208] In a practical implementation, e.g. a portable device, such as hearing aid, we prefer to avoid the division and reduce the number of computations, so we propose the following log normalized covariance measure

ρ(m)=Σ.sub.k{log(max{0Im{Ĉ.sub.x12}+1})−log(Ĉ.sub.x11)},   (14)

[0209] Two instances of the (log) normalized covariance measure are calculated, a fast instance {tilde over (ρ)}(m) and an instance ρ(m) with variable update rate. The fast instance {tilde over (ρ)}(m) is based on the fast variance estimate

[00016]$\begin{array}{cc}{\tilde{C}}_{x.Math..Math.11}\left(m\right)=\{\begin{array}{c}\tilde{\alpha }.Math.{\tilde{C}}_{x.Math..Math.11}\left(m-1\right)+\left(1-\tilde{\alpha }\right).Math.X\left(m\right).Math.{X\left(m\right)}^H,\mathrm{Target}.Math..Math.\mathrm{present}\\ {\tilde{C}}_{x.Math..Math.11}\left(m-1\right),\mathrm{Target}.Math..Math.\mathrm{absent}\end{array};& \left(15\right)\end{array}$

where {tilde over (α)} is a fast time constant smoothing factor, and the corresponding fast covariance estimate

[00017]$\begin{array}{cc}{\tilde{C}}_{x.Math..Math.12}\left(m\right)=\{\begin{array}{c}\tilde{\alpha }.Math.{\tilde{C}}_{x.Math..Math.12}\left(m-1\right)+\left(1-\tilde{\alpha }\right).Math.X\left(m\right).Math.{X\left(m\right)}^H,\mathrm{Target}.Math..Math.\mathrm{present}\\ {\tilde{C}}_{x.Math..Math.12}\left(m-1\right),\mathrm{Target}.Math..Math.\mathrm{absent}\end{array};& \left(16\right)\end{array}$

according to

ρ(m)=Σ.sub.k{log(max{0, Im{{tilde over (C)}.sub.x12}+1})−log({tilde over (C)}.sub.x11)},   (17)

[0210] Similar expressions for the instance with variable update rate ρ(m), based on equivalent estimators C.sub.x11(m) and C.sub.x12(m) using a variable smoothing factor α(m) can be written:

[00018]$\begin{array}{cc}{\stackrel{\_}{C}}_{x.Math..Math.11}\left(m\right)=\{\begin{array}{c}\stackrel{\_}{\alpha }.Math.{\stackrel{\_}{C}}_{x.Math..Math.11}\left(m-1\right)+\left(1-\stackrel{\_}{\alpha }\right).Math.X\left(m\right).Math.{X\left(m\right)}^H,\mathrm{Target}.Math..Math.\mathrm{present}\\ {\stackrel{\_}{C}}_{x.Math..Math.11}\left(m-1\right),\mathrm{Target}.Math..Math.\mathrm{absent}\end{array};& \left({15}^{\prime }\right)\end{array}$

where {tilde over (α)} is a fast time constant smoothing factor, and the corresponding fast covariance estimate

[00019]$\begin{array}{cc}{\stackrel{\_}{C}}_{x.Math..Math.12}\left(m\right)=\{\begin{array}{c}\stackrel{\_}{\alpha }.Math.{\stackrel{\_}{C}}_{x.Math..Math.12}\left(m-1\right)+\left(1-\stackrel{\_}{\alpha }\right).Math.X\left(m\right).Math.{X\left(m\right)}^H,\mathrm{Target}.Math..Math.\mathrm{present}\\ {\stackrel{\_}{C}}_{x.Math..Math.12}\left(m-1\right),\mathrm{Target}.Math..Math.\mathrm{absent}\end{array};& \left({16}^{\prime }\right)\end{array}$

according to

ρ(m)=Σ.sub.k{log(max{0, Im{C.sub.x12}+1})−log(C.sub.x11)},   (17′)

[0211] The smoothing factor α of the variable estimator is changed to fast when the normalized covariance measure of the variable estimator deviates too much from the normalized covariance measure of the variable estimator, otherwise the smoothing factor is slow, i.e.

[00020]$\begin{array}{cc}\stackrel{\_}{\alpha }\left(m\right)=\{\begin{array}{c}{\alpha }_0,.Math.\tilde{\rho }\left(m\right)-\stackrel{\_}{\rho }\left(m\right).Math.\leqq \epsilon \\ \tilde{\alpha },.Math.\tilde{\rho }\left(m\right)-\stackrel{\_}{\rho }\left(m\right).Math.>\epsilon \end{array}& \left(18\right)\end{array}$

where α.sub.0 is a slow time constant smoothing factor, i.e. α.sub.0<α, and ∈ is a constant. Note that the same smoothing factor α(m) is used across frequency bands k.

[0212] FIGS. 15A, 15B and 15C illustrate a general embodiment of the variable time constant covariance estimator as outlined above.

[0213] FIG. 15A schematically shows a covariance smoothing unit according to the present disclosure. The covariance unit comprises a pre-smoothing unit (PreS) and a variable smoothing unit (VarS). The pre-smoothing unit (PreS) makes an initial smoothing over time of instantaneous covariance matrices C(m)=X(m)X(m).sup.H (e.g. representing the covariance/variance of noisy input signals X) in K frequency bands and provides pre-smoothed covariance matrix estimates X.sub.11, X.sub.12 and X.sub.22 (<C>.sub.pre=<X(m)X(m).sup.H>, where <•> indicates LP-smoothing over time). The variable smoothing unit (VarS) makes a variable smoothing of the signals X.sub.11, X.sub.12 and X.sub.22 based on adaptively determined attack and release times in dependence of changes in the acoustic environment as outlined above, and provides smoothed covariance estimators C.sub.x11(m), C.sub.x12(m), and C.sub.x22(m).

[0214] The pre-smoothing unit (PreS) makes an initial smoothing over time (illustrated by ABS-squared units |•|.sup.2 for providing magnitude squared of the input signals X.sub.i(k,m) and subsequent low-pass filtering provided by low-pass filters LP) to provide pre-smoothed covariance estimates C.sub.x11, C.sub.x12 and C.sub.x22, as illustrated in FIG. 15B. X.sub.1 and X.sub.2 may e.g. represent first (e.g. front) and second (e.g. rear) (typically noisy) microphone signals of a hearing aid. Elements C.sub.x11, and C.sub.x22, represent variances (e.g. variations in amplitude of the input signals), whereas element C.sub.x12 represent co-variances (e.g. representative of changes in phase (and thus direction) (and amplitude)).

[0215] FIG. 15C shows an embodiment of the variable smoothing unit (VarS) providing adaptively smoothed of covariance estimators C.sub.x11(m), C.sub.x12(m), and C.sub.x22(m), as discussed above.

[0216] The Target Present input is e.g. a control input from a voice activity detector. In an embodiment, the Target Present input (cf. signal TP in FIG. 15A) is a binary estimate (e.g. 1 or 0) of the presence of speech in a given time frame or time segment. In an embodiment, the Target Present input represents a probability of the presence (or absence) of speech in a current input signal (e.g. one of the microphone signals, e.g. X.sub.1(k,m)). In the latter case, the Target Present input may take on values in the interval between 0 and 1. The Target Present input may e.g. be an output from a voice activity detector (cf. VAD in FIG. 15C), e.g. as known in the art.

[0217] The Fast Rel Coef, the Fast Atk Coref, the Slow Rel Coef, and the Slow Atk Coef are fixed (e.g. determined in advance of the use of the procedure) fast and slow attack and release times, respectively. Generally, fast attack and release times are shorter than slow attack and release times. In an embodiment, the time constants (cf. signals TC in FIG. 15A) are stored in a memory of the hearing aid (cf. e.g. MEM in FIG. 15A). In an embodiment the time constants may be updated during use of the hearing aid.

[0218] It should be noted that the goal of the computation of y=log(max(Im{x12}+1,0))−log(x11) (cf. two instances in the right part of FIG. 15C forming part of the determination of the smoothing factor α(m)) is to detect changes in the acoustical sound scene, e.g. sudden changes in target direction (e.g. due to a switch of current talker in discussion/convesation). The exemplary implementation in FIG. 15C is chosen for its computational simplicity (which is of importance in a hearing device having a limited power budget), as provided by the conversion to a logarithmic domain. A mathematically more corect (but computationally more complex) implementation would be to compute y=x12/x11 (as exemplified in the determination of β illustrated in FIG. 3 and FIG. 7 (and FIG. FIG. 17A, 17B).

[0219] The adaptive low-pass filters used in FIG. 15C can e.g. be implemented as shown in FIG. 4, where coef is the smoothing factor α(m) (or {tilde over (α)}(m)).

[0220] FIGS. 16A, 16B and 16C illustrate a particular embodiment of the variable time constant covariance estimator as outlined above. The difference of the embodiment of FIGS. 16A, 16B and 16C to the general embodiment of FIG. 15A, 15B, 15C is that the inputs are beamformed signals formed by beam patterns C1 and C2 (instead of microphone signals x directly). FIG. 16D schematically illustrates the determination of β based on smoothed covariance matrices (<|C2 |.sup.2>, <C1C2*>) according to the present disclosure (as exemplified in FIG. 17A, 17B).

[0221] The above scheme may e.g. be relevant for adaptively estimating a direction of arrival of alternatingly active sound sources at different locations (e.g. at different angles in a horizontal plane relative to a user wearing one or more hearing aids according to the present disclosure).

[0222] FIG. 17A corresponds to FIGS. 3 and FIG. 17B corresponds to FIG. 7, but in FIGS. 17A and 17B, the variable time constant covariance estimator according to the present disclosure (and as depicted in FIG. 16A-16C) is used for adaptively smoothing β.

[0223] FIG. 18 comprises a pre-smoothing unit (PreS), a variable smoothing unit (VarS) and a β calculation unit (beta) as also illustrated in FIGS. 17A and 17B, but in an alternative embodiment.

[0224] FIG. 18 illustrates how β can be determined from the (e.g. smoothed) noise covariance matrix <C.sub.v>(during speech pauses ‘VAD=0’) according to the present disclosure, contrary to calculating the beamformers. The LP blocks may be time varying (e.g. adaptive) as e.g. shown in connection with FIGS. 15C and FIG. 16C. Instead of showing all the multiplications, two matrix multiplication blocks (NUMC, and DENC, respectively), for determining the numerator (num) and denominator (den) of the calculation of β are indicated in FIG. 18. An advantage of this implementation is that the beamformer coefficients may be modified without affecting the smoothing. This comes at the cost that this implementation requires more multiplications and an additional LP filter.

[0225] It is intended that the structural features of the devices described above, either in the detailed description and/or in the claims, may be combined with steps of the method, when appropriately substituted by a corresponding process.

[0226] As used, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well (i.e. to have the meaning “at least one”), unless expressly stated otherwise. It will be further understood that the terms “includes,” “comprises,” “including,” and/or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It will also be understood that when an element is referred to as being “connected” or “coupled” to another element, it can be directly connected or coupled to the other element but an intervening elements may also be present, unless expressly stated otherwise. Furthermore, “connected” or “coupled” as used herein may include wirelessly connected or coupled. As used herein, the term “and/or” includes any and all combinations of one or more of the associated listed items. The steps of any disclosed method is not limited to the exact order stated herein, unless expressly stated otherwise.

[0227] 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.

[0228] The claims are 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.

[0229] Accordingly, the scope should be judged in terms of the claims that follow.