Feedback cancellation in hearing aid devices using filter tap coherence values

Abstract

The system (20) for hearing assistance includes one or more microphones (23, 24), a speaker (28), and processing circuitry (26). The one or more microphones are mounted near a subject's head and configured to output electrical signals in response to acoustic waves incident on the microphones. The speaker is configured to be mounted near the subject's ears. The processing circuitry amplifies and filters the electrical signals to generate drive signals for input to the speakers using a digital filter (100) having multiple taps, with respective tap coefficients selected to suppress feedback from the speaker to the microphones, and is configured to adaptively calculate tap coefficients, estimating coherence values ​​for each of the tap coefficients over time and weighting updates applied to the tap coefficients in response to the respective coherence values.

Description

【Technical Field】 【0001】 The present invention generally relates to hearing aids, and more particularly to devices and methods for acoustic feedback cancellation. 【Background Art】 【0002】 Speech understanding in noisy environments is a significant problem for hearing-impaired individuals. Hearing impairment typically involves, in addition to gain loss, a reduction in the temporal resolution of the sensory system. These characteristics further degrade the ability of hearing-impaired individuals to filter the target sound source from background noise, particularly the ability to understand speech in noisy environments. 【0003】 Some new hearing aids have a directional hearing mode that improves speech intelligibility in noisy environments. This mode uses an array of microphones and applies beamforming technology to combine multiple microphone inputs into a single directional audio output channel. The output channel has spatial characteristics that increase the contribution of acoustic waves arriving from the target direction compared to the contribution of acoustic waves from other directions. 【0004】 For example, PCT International Patent Application Publication No. 2017 / 158507, the disclosure of which is incorporated herein by reference, describes a hearing aid device that includes a case configured to be physically fixed to a mobile phone. The array of microphones is spaced within the case and is configured to generate an electrical signal in response to an acoustic input to the microphones. The interface is fixed within the case together with a processing circuit configuration that is coupled to receive and process the electrical signals from the microphones and generate a combined signal for output via the interface. 【0005】 As another embodiment, PCT International Patent Application Publication 2021 / 074818, whose disclosure is incorporated herein by reference, describes a hearing aid comprising an eyeglass frame including a front piece and temples, where one or more microphones are mounted at first positions on each of the front pieces and configured to output electrical signals in response to first acoustic waves incident on the microphones. A speaker mounted at a second position on one of the temples outputs a second acoustic wave. A processing circuit configuration generates a drive signal for the speaker by processing the electrical signals output by the microphones, causing the speaker to reproduce a selected sound generated by the first acoustic wave from the first position to the second position with equal delays within 20% of the transit time of the first acoustic wave, thereby creating constructive interference between the first and second acoustic waves. [Overview of the project] [Problems that the invention aims to solve] 【0006】 The embodiments of the present invention described below provide improved devices and methods for hearing assistance. [Means for solving the problem] 【0007】 One embodiment described herein provides a system for hearing assistance comprising one or more microphones, a speaker, and a processing circuit configuration. The one or more microphones are mounted near the head of a subject and are configured to output electrical signals in response to acoustic waves incident on the microphones. The speaker is configured to be mounted near the ear of a subject. The processing circuit configuration is configured to amplify and filter the electrical signals to generate a drive signal for input to the speaker, using a digital filter having multiple taps, each having a tap coefficient selected to suppress feedback from the speaker to the microphones, and to adaptively calculate the tap coefficients while estimating the coherence value of each tap coefficient over time and weighting updates applied to the tap coefficients according to each coherence value. 【0008】 In some embodiments, the processing circuit configuration is configured to adapt the tap coefficients to estimate the transfer function between the speaker and one or more microphones. In other embodiments, the processing circuit configuration is configured to adapt the tap coefficients using a gradient descent method with respective convergence coefficients. In yet another embodiment, the processing circuit configuration is configured to calculate the convergence coefficients based on coherence values. 【0009】 In one embodiment, the processing circuit configuration is configured to calculate the convergence coefficient by multiplying a common convergence coefficient by the respective coherence values. In another embodiment, the processing circuit configuration is configured to evaluate the coherence value of a given tap based on a plurality of coefficient updates calculated for a given tap over a specified period. In yet another embodiment, the hearing system includes an eyeglass frame, and a microphone and a speaker are mounted at their respective positions on the eyeglass frame. 【0010】 In some embodiments, one or more microphones include multiple microphones, and the processing circuit configuration is configured to apply beamforming functionality to the electrical signals output by the multiple microphones in order to emphasize selected sounds occurring within a selected angular range and suppress background noise occurring outside the selected angular range. 【0011】 Furthermore, according to one embodiment described herein, a method for hearing assistance is provided, comprising mounting an array of microphones that output electrical signals in response to acoustic waves incident on the microphones near the head of a subject, and mounting speakers near the ears of the subject. The electrical signals are amplified and filtered to generate a drive signal for input to the speakers using a digital filter having multiple taps, each having a tap coefficient selected to suppress feedback from the speakers to the microphones. The tap coefficients are calculated adaptively, while the coherence value of each tap coefficient is estimated over time, and updates applied to the tap coefficients are weighted according to the respective coherence values. 【0012】 Furthermore, according to another embodiment described herein, a head-mounted device (HMD) is provided comprising a frame, one or more microphones, a speaker, and a processing circuit configuration. The frame is configured to be attached to the head of a subject. One or more microphones are attached to the frame and configured to output electrical signals in response to acoustic waves incident on the microphones. A speaker is attached to the frame. The processing circuit configuration is configured to amplify and filter the electrical signals to generate a drive signal for input to the speaker, using a digital filter having multiple taps, each having a tap coefficient selected to suppress feedback from the speaker to the microphones, and to adaptively calculate the tap coefficients while estimating the coherence value of each tap coefficient over time and weighting updates applied to the tap coefficients according to each coherence value. 【0013】 In some embodiments, the HMD includes a device selected from a list including eyewear devices, glasses, eyeglass frames, goggles, helmets, visors, headsets, and clip-on devices. In other embodiments, one or more microphones are mounted on the front piece of the frame, and speakers are mounted on the frame near the subject's ears. 【0014】 The present invention will be better understood by reading the following detailed description of embodiments in conjunction with the drawings. [Brief explanation of the drawing] 【0015】 [Figure 1] This is a schematic diagram illustrating a hearing aid device based on an eyeglass frame according to one embodiment of the present invention. [Figure 2] This is a block diagram illustrating the details of a hearing aid device according to one embodiment of the present invention. [Figure 3] This is a block diagram illustrating the details of a feedback canceller applicable to a hearing aid device according to one embodiment of the present invention. [Figure 4A] This block diagram schematically illustrates a processing scheme that supports both beamforming and feedback cancellation according to an embodiment of the present invention. [Figure 4B] This block diagram schematically illustrates a processing scheme that supports both beamforming and feedback cancellation according to an embodiment of the present invention. [Modes for carrying out the invention] 【0016】 Overview Despite the need for directional hearing aids and the theoretical advantages of microphone arrays in this regard, in practice, the directional performance of hearing aids falls far short of the performance achieved by natural hearing. Generally, good directional hearing aids require a relatively large number of microphones placed at sufficient intervals in an inconspicuous design, while still allowing the user to easily direct the hearing aid's directional response to a point of interest, such as a conversation partner in a noisy environment. The processing circuit configuration applies beamforming filters to the signals output by the microphones in response to the incident acoustic waves, suppressing background noise and generating an audio output that emphasizes the sound hitting the microphone array within an angular range around the direction of interest. The audio output needs to reproduce the most natural auditory experience possible while minimizing annoying artifacts. 【0017】 One of these artifacts is a loud whistling sound that can occur due to acoustic feedback from the audio output of a speaker located near the user's ear to the microphone input. Such a whistling sound occurs when the acoustic feedback gain of the hearing aid at a given frequency exceeds a certain threshold. Feedback cancellation in hearing aid devices is generally more difficult than in applications such as video conferencing and telephone calls, where the echo signal can be delayed by about 100 milliseconds, but in hearing aid devices, the feedback signal is usually delayed by less than 20 milliseconds, resulting in a higher correlation between the spectrum of the system output and input. Conventional solutions to suppress or cancel the feedback signal include lowering the hearing aid gain and filtering the range of audio frequencies in which the feedback occurs, but these solutions also reduce the effectiveness of the hearing aid in amplifying faint sounds and high frequencies. It is also possible to mechanically reduce the feedback gain by fitting an ear mold to the user's ear, but many users find this solution uncomfortable and unsightly. 【0018】 The embodiments of the invention described herein address the problem of acoustic feedback by providing a method and system for novel feedback cancellation by estimating a feedback signal and subtracting the feedback signal from an input signal. In the disclosed embodiments, an array of microphones mounted near a user's head outputs an electrical signal in response to incident acoustic waves impinging on the microphones. A speaker is mounted near the user's ear. The processing circuitry amplifies and filters the electrical signal to generate a drive signal for input to the speaker using a digital filter having a plurality of taps with respective tap coefficients selected to suppress feedback from the speaker to the microphone, estimates the respective coherence values of the tap coefficients over time, and adaptively calculates the tap coefficients while weighting updates applied to the tap coefficients according to the respective coherence values. 【0019】 In some embodiments, the microphone and the speaker are attached to a frame attached to the user's head. In some of the embodiments described below, the microphone and the speaker are attached to a眼镜frame. Alternatively, the microphone and the speaker can be attached to other types of frames or head-mounted devices (HMDs) such as virtual reality (VR) or augmented reality (AR) headsets, or to other types of attachment configurations. 【0020】 In this context, an HMD comprises any type of frame to which a microphone and a speaker can be attached. The HMD can be selected from a list comprising, but not limited to, eyewear devices, glasses, eyeglass frames, goggles, helmets, visors, headsets, and clip-on devices. In some embodiments, one or more microphones are attached to a front piece of the frame and the speaker is attached to the frame near the subject's ear. 【0021】 It should be noted that the Chinese character "眼镜" in the original text seems to be incorrect. It might be intended to be "眼镜" which means "eyeglasses" in English. I translated it as "eyeglass" in the above translation. If this is not what you meant, please correct the original text and I will provide a more accurate translation.In some embodiments, the processing circuit configuration adapts the tap coefficients to estimate the transfer function between a speaker and one or more microphones. The processing circuit configuration uses the estimated transfer function to estimate a feedback signal that is subtracted from the input signal. The processing circuit configuration may adapt the tap coefficients using a gradient descent method having respective convergence coefficients, and the processing circuit configuration calculates the respective convergence coefficients based on a coherence value. In one embodiment, the processing circuit configuration calculates the convergence coefficients by multiplying a common convergence coefficient by each respective coherence value. 【0022】 In some embodiments, the processing circuit configuration evaluates the coherence value of a given tap based on a plurality of coefficient updates calculated for the given tap over a specified period. 【0023】 In some embodiments, the system includes a pair of glasses, and the microphone and the speaker are attached to respective positions on the pair of glasses. 【0024】 In one embodiment, the one or more microphones include a plurality of microphones, and the processing circuit configuration applies a beamforming function to the electrical signals output by the plurality of microphones to enhance selected sounds generated within a selected angular range and suppress background sounds generated outside the selected angular range. 【0025】 System description Figure 1 is a schematic diagram of a hearing aid device 20 integrated into an eyeglass frame 22 according to one embodiment of the present invention. An array of microphones 23, 24 is mounted at respective positions on the eyeglass frame 22 and outputs an electrical signal in response to acoustic waves incident on the microphones. In the illustrated embodiment, microphone 23 is mounted on the front piece 30 of the frame 22, while microphone 24 is mounted on temples 32 connected to the respective ends of the front piece 30. While the broad array of microphones 23, 24 shown in Figure 1 is useful in some applications of the present invention, the signal processing and hearing aid principles described herein can instead be applied using fewer microphones with necessary modifications. For example, these principles can be applied using an array of microphones 23 on the front piece 30, and can also be applied to devices using other microphone mounting configurations that are not necessarily eyeglass-based. 【0026】 The processing circuit configuration 26 is fixed within or connected to the spectacle frame 22 and coupled by electrical wiring 27, such as traces on a flexible printed circuit, to receive electrical signals output from microphones 23, 24. While Figure 1 illustrates the processing circuit configuration 26, for simplification, at certain positions on the temples 32, some or all of the processing circuit configuration may instead be located on the front piece 30 or in a unit externally connected to the frame 22. The processing circuit configuration 26 mixes the signals from the microphones and, for example, applies beamforming functionality to enhance sounds occurring within a selected angular range and suppress background noise occurring outside that range, thereby producing an audio output with a specific directional response. Typically, the directional response coincides with the angular orientation of the frame 22, though not necessarily. In addition, the processing circuit configuration suppresses acoustic signals picked up by the microphones and emitted from the speakers. 【0027】 These signal processing functions of the processing circuit configuration 26 will be described in more detail below. 【0028】 The processing circuit configuration 26 can transmit audio output to the user's ears via any suitable type of interface and speaker. In the illustrated embodiment, the audio output is generated by a drive signal to drive one or more audio speakers 28, typically mounted on the temples 32 near the user's ears. Although only one speaker 28 is illustrated for each temple 32 in Figure 1, the device 20 may instead have only a single speaker on one of the temples 32, or two or more speakers mounted on one or both of the temples 32. In the latter case, the processing circuit configuration 26 may apply a beamforming function to the drive signal so that the acoustic waves from the speakers are directed towards the user's ears. Alternatively, the drive signal may be transmitted to a speaker inserted into the ear, or it may be transmitted wirelessly, for example, as a magnetic signal, to a telecoil in a hearing aid (not shown) worn by the user with the eyeglass frame. 【0029】 Signal processing Figure 2 is a schematic block diagram illustrating the details of a processing circuit configuration 26 in a hearing aid device 20 according to one embodiment of the present invention. The processing circuit configuration 26 can be implemented on a single integrated circuit chip, or the functions of the processing circuit configuration 26 can be distributed across multiple chips that may be located inside or outside the eyeglass frame 22. Although Figure 2 illustrates one particular implementation, the processing circuit configuration 26 may comprise any suitable combination of analog and digital hardware circuits, along with a suitable interface for receiving electrical signals output by microphones 23, 24 and outputting drive signals to the speaker 28. 【0030】 In this embodiment, the microphones 23 and 24 are equipped with an integrated analog-to-digital converter and output a digital audio signal to the processing circuit configuration 26. Alternatively, the processing circuit configuration 26 may be equipped with an analog-to-digital converter that converts the analog output of the microphones into a digital format. The processing circuit configuration 26 typically includes a suitable programmable logic component 40, such as a digital signal processor (DSP) or gate array, which performs the necessary filtering, mixing, and feedback cancellation functions to generate and output the drive signal for the speaker 28 in digital format. 【0031】 These filtering and mixing functions typically involve the application of a beamforming filter 42 having coefficients selected to create a desired directional response. Specifically, in some embodiments, the coefficients of the beamforming filter 42 are calculated to enhance the sound hitting the frame 22 (and thus the microphones 23, 24) within a selected angular range. Further details of filters that may be used for beamforming purposes are described below. 【0032】 Alternatively or additionally, the processing circuit configuration 26 may include a neural network (not shown) trained to determine and apply coefficients used in the beamforming filter 42. Further alternative or additionally, the processing circuit configuration 26 may include a microprocessor programmed by software or firmware to perform at least some of the functions described herein. 【0033】 The processing circuit configuration 26 can apply any suitable beamforming function known in the art in either the time domain or the frequency domain when implementing the beamforming filter 42. Beamforming algorithms that can be used in this context are described, for example, in the aforementioned PCT International Patent Application Publication No. 2017 / 158507 (especially pages 10-11) and U.S. Patent No. 10,567,888 (especially paragraph 9). 【0034】 In one embodiment, the processing circuit configuration 26 applies a minimum dispersion-free-distortion (MVDR) beamforming algorithm when deriving the coefficients of the beamforming filter 42. This type of algorithm is advantageous in achieving fine spatial resolution and distinguishing between sounds originating from the direction of interest and sounds originating from the user's own voice. The MVDR algorithm maximizes the signal-to-noise ratio (SNR) of the audio output by minimizing the average energy (while keeping target distortion low). This algorithm can be implemented in the frequency domain by calculating a complex weight vector F(ω) of the output signal from each microphone at each frequency, which is expressed by the following equation. 【0035】 【number】 In this equation, W(ω) is the propagation delay vector between microphones 23, and the desired response of the beamforming filter is expressed as a function of angle and frequency, S zz (ω) is the cross-spectral density matrix, which represents the covariance of the acoustic signal in the time-frequency domain. To calculate the coefficients of the beamforming filter 42, S zz (ω) is measured or calculated for isotropic far-field noise. 【0036】 In an alternative embodiment, the processing circuit configuration 26 applies a linear-constrained minimum variance (LCMV) algorithm when deriving the coefficients of the beamforming filter 42. LCMV beamforming allows the beamforming filter to pass a signal from a desired direction with a specified gain and phase delay, while minimizing power from interference signals and noise from all other directions. 【0037】 In some embodiments, the processing circuit configuration 26 includes a feedback canceller 44 to suppress acoustic feedback from the speaker to the microphone. For this purpose, the feedback canceller 44 uses a digital filter (not shown) having multiple taps, each having a tap coefficient selected to suppress feedback from the speaker to the microphone, to estimate the coherence value of each tap coefficient over time, and adaptively calculates the tap coefficients while weighting the updates applied to the tap coefficients according to each coherence value. The feedback canceller is described in detail below with reference to Figure 3. 【0038】 The audio output circuit 46, for example, includes a suitable codec and digital-to-analog converter to convert the digital drive signal output from the beamforming filter 42 (or from the feedback canceller 44 following the beamforming filter) into an analog format. The analog filter 48 further performs filtering and analog amplification functions to optimize the analog drive signal to the speaker 28. 【0039】 A control circuit 50, such as an embedded microcontroller, controls the programmable functions and parameters of a processing circuit configuration 26, which may include a feedback canceller 44. A communication interface 52, such as Bluetooth(R) or another wireless interface, allows the user and / or hearing professional to set and adjust these parameters as needed. A power circuit 54, such as a battery inserted in the stalk 32, supplies power to the other components of the processing circuit configuration. 【0040】 Feedback cancellation process As mentioned above, sound waves generated by the speaker of a hearing aid device can be picked up by the microphone of the hearing aid device, sometimes resulting in a whistling or feedback sound. The purpose of a feedback canceller is to prevent the whistling artifact by reducing the amount of feedback signal in the signal generated by the microphone. 【0041】 Next, the principle of feedback cancellation is explained. Out(t) represents the signal output by the hearing aid device, p(t) represents the signal received by the microphone from the output of the hearing aid device only (the version of Out(t) received by the microphone), and y(t) represents the signal received by the microphone from all audio sources other than the speaker of the hearing aid device, where t represents the time axis. The overall signal x(t) generated by the microphone is given by x(t) = y(t) + p(t). 【0042】 The feedback canceller generates a feedback signal based on the output signal Out(t-Δt) (e.g., a reference signal) generated before a period of Δt, as follows: 【number】 To estimate, the feedback canceller is 【number】 The transfer function from the hearing aid device output (speaker) to the microphone is estimated, and the estimated transfer function is applied to the signal Out(t-Δt). 【number】 It generates an estimated feedback signal given by [the specified method]. 【0043】 The feedback canceller further subtracts the estimated feedback signal from x(t), 【number】 This generates the signal x'(t) given by [the function]. Here, feedback is suppressed. In digital form, the transfer function 【number】 This can be carried out using an adaptive filter with multiple taps, and the tap coefficients are adapted using any suitable adaptation method. The tap coefficients can be adapted using any suitable gradient descent method, such as least mean squares (LMS) or normalized LMS (NLMS). Alternatively, other suitable adaptation methods can also be used. As described below with reference to Figures 4A and 4B, depending on whether the feedback cancellation is performed before or after beamforming, the adaptive filter can model the transfer function between a speaker and multiple microphones, or between a speaker and a single microphone. 【0044】 Figure 3 is a schematic block diagram showing details of a feedback canceller 44 applicable to a hearing aid device 20 according to one embodiment of the present invention. Alternatively, the principle of this feedback canceller may also be applied to other devices and systems equipped with a suitable microphone array, speaker, and signal processing function. 【0045】 The feedback canceller 44 in Figure 3 implements the feedback cancellation principle described above in a digital form, where various signals are sampled on a digital time axis denoted as 'n'. In the embodiment of Figure 3, the feedback canceller 44 is received by microphones 23 and 24 and receives an input signal x(n) containing the feedback signal from speaker 28. The feedback canceller uses a subtractor 104 to estimate the feedback signal from x(n). 【0046】 【number】 Subtracting this value generates a signal x'(n) where the feedback is suppressed or canceled. 【0047】 The feedback canceller 44 is an adaptive filter having N taps, each with its own tap coefficient. 【number】 The adaptive filter 100 comprises 100, where N is an integer greater than 1. The feedback canceller generates an estimated feedback signal by filtering the output signal Out(n) using the current values ​​of the tap coefficients of the adaptive filter 100. In some embodiments, the output signal Out(n) comprises the drive signal to the speaker in digital form. The adaptive filter may have any suitable number N taps. In exemplary embodiments, the number of taps is 100 or more, for example, around 120 taps. The main reasons for selecting so many taps are (i) that the feedback cancellation is performed with a tight beamformer, and (ii) that the frequency response of the speaker in the underlying hearing eyewear is significantly different from a flat frequency response. For these reasons, the processed signal becomes blurred over a relatively long period of time, requiring a relatively long filter. 【0048】 The tap adapter 108 updates the tap coefficients of the adaptive filter 100 using any suitable gradient descent method, such as the LMS method or the NLMS method. Δh(n) represents the coefficient update vector corresponding to each tap of the adaptive filter 100. The vector Δh(n) has the same length N as the adaptive filter 100. In this embodiment, the tap adapter is: 【0049】 【number】 Execute sequential update steps as given by [the specified method]. 【0050】 Here, the update vector Δh(n) is: Δh(n) = μ·Out(n)·X(n) It is given by. 【0051】 μ is the scalar convergence coefficient of the underlying gradient descent method, and the vector X(n) is, X(n) = [x(n-N+1)...x(n)] It is given by. 【0052】 Next, embodiments are described in which the adaptation of tap coefficients by the tap adapter 108 is based on multiple convergence coefficients rather than a single scalar convergence coefficient. In such embodiments, for each tap, a common convergence coefficient μ is weighted by its respective weight value. In some embodiments, the tap adapter calculates the weight value by calculating the respective tap coherence value, as described herein. This approach provides a time-based weighting mechanism for modifying the update Δh(n) applied to the tap coefficients of the adaptive filter. The inventors have found that, for example, in open-ear hearing eyewear, weighting the update of tap coefficients by the time coherence value of each tap can significantly improve feedback cancellation performance. 【0053】 The performance of a feedback cancellation method can be determined, for example, by measuring the maximum acoustic output gain at which the underlying system remains stable without whistling. The inventors found that the applicable gain using the disclosed coherence-based feedback cancellation method is significantly higher than the gain achievable when the coherence value is omitted. 【0054】 Generally, using a coherence value involves evaluating the updates that are adaptively applied to each tap of an adaptive filter over a short period, for example, 16 milliseconds (or any other suitable period), and weighting each tap coefficient update based on the coherence value. In some embodiments, a coherence value C is used to weight the coefficient update of the i-th tap. i For example, 【0055】 【number】 It is given by. 【0056】 Here, n represents the digital time index, 【number】 `x` represents the coefficient update applied to the i-th tap at time n, and `W` represents the number of samples used to calculate the coherence value. The coherence value ranges from 0 to 1, taking the maximum value of 1 if all coefficient values ​​used for its calculation are equal to each other. Although not required, the coefficient value may be calculated based on a sequence of W consecutive tap updates most recently applied to the relevant tap. Alternatively, other recent W tap updates can also be used. Gradient coefficient of the i-th tap coefficient weighted by the i-th coherence value. 【0057】 【number】 teeth, 【number】 It is given by. 【0058】 Coherence value C i This indicates the confidence level associated with each coefficient update. The gradient coefficient μ is weighted higher when the tap is associated with a large coherence value (considered to be a high-confidence update) and lower when the tap is associated with a small coherence value (considered to be a low-confidence update). 【0059】 The method for calculating the above coherence value is given as an example, and other types of coherence values ​​can also be used. For example, a less complex sign coherence value is: 【number】 It is given by. 【0060】 As another example, the coherence value is each coherence value C i to, 【number】 The coefficient C is given by g This can be generalized by multiplying by . 【0061】 Here, the sum of the last equation is taken over the number of taps W'>1. 【0062】 In embodiments where feedback cancellation is performed in the frequency domain, a phase coherence coefficient can be applied. An exemplary formulation of this type can be found, for example, in the paper “Phase Coherence Imaging: Principles, applications and current developments,” Bruges, Belgium, Signal Processing in Acoustics: PSP (2 / 3) Presentation 1. 【0063】 Beamforming and feedback cancellation scheme Figures 4A and 4B are schematic block diagrams illustrating a processing scheme that supports both beamforming and feedback cancellation according to an embodiment of the present invention. 【0064】 The schemes in Figures 4A and 4B differ in the order in which beamforming and feedback cancellation are performed. 【0065】 In the scheme of Figure 4A, input signals from multiple microphones are processed by a beamforming filter, such as the beamforming filter 42 in Figure 2. The signal output by the beamforming filter is then feedback-cancelled, for example, using the feedback canceller 44 in Figures 2 and 3. In such embodiments, the adaptive filter in Figure 3... 【0066】 【number】 This models the transfer function from a speaker (e.g., 28 in Figure 2) to a composite (virtual) microphone comprising multiple microphones. Interface 120 provides the speaker with the signal output by a feedback canceller. Interface 120 may include, for example, a codec / DAC (e.g., 46 in Figure 2) followed by an analog filter (e.g., 48 in Figure 2). 【0067】 In the scheme shown in Figure 4B, input signals from multiple microphones are each processed by a dedicated feedback canceller 44. The output of the feedback canceller is fed into a beamforming filter 42, whose output is provided to the speaker via interface 120. In such an embodiment, the adaptive filter shown in Figure 3 【0068】 【number】 This models the transfer function from the speaker (for example, 28 in Figure 2) to each individual microphone. 【0069】 The scheme in Figure 4A is less complex than the scheme in Figure 4B because it uses only one feedback canceller instead of multiple feedback cancellers. Furthermore, applying separate feedback cancellation to individual microphones (as in Figure 4B) can reduce the inter-microphone correlation necessary for proper beamforming filter operation, so the beamforming performance in the scheme in Figure 4A may be better than that in Figure 4B. 【0070】 On the other hand, the scheme in Figure 4B may be more advantageous than the scheme in Figure 4A because more information and degrees of freedom are available to mitigate problems such as howling when feedback cancellation is performed individually on each microphone. Furthermore, providing multiple microphone feedback-free audio channels (as in Figure 4B) can be used when implementing various algorithms other than beamforming. Exemplary relevant algorithms in this regard include (but are not limited to) self-voice detection, direction of arrival estimation, transfer function, and measurement of room sound levels. 【0071】 While the embodiments described herein primarily address feedback cancellation in hearing aid devices, the methods and systems described herein can also be used for feedback cancellation in other HMD devices and in other applications such as noise-canceling headphones. 【0072】 The embodiments described above are given as examples, and it will be recognized that the following claims are not particularly limited to those illustrated and described above. Rather, the scope includes both combinations and partial combinations of the various features described above, as well as variations and modifications thereof not disclosed in the prior art, which a person skilled in the art would likely conceive of upon reading the foregoing description. Documents incorporated by reference in this patent application shall be considered integral parts thereof, however, with respect to terms defined in such incorporated documents in a manner that contradicts the definitions made expressly or implicitly in this specification, only the definitions provided herein shall be considered.

Claims

[Claim 1] A hearing aid system (20), One or more microphones (23, 24) are mounted near the subject's head and configured to output electrical signals in response to acoustic waves incident on the microphones, A speaker (28) is configured to be attached near the ear of the subject, The system comprises a processing circuit configuration (26) configured to amplify and filter the electrical signal to generate a drive signal for input to the speaker using a digital filter (100) having a plurality of taps, each having a tap coefficient selected to suppress feedback from the speaker to the microphone, and configured to estimate the coherence value of each of the tap coefficients over time and to adaptively calculate the tap coefficients while weighting the updates applied to the tap coefficients according to the respective coherence values, In order to adaptively calculate the tap coefficients, the processing circuit configuration is configured to adapt the tap coefficients using a gradient descent method having respective convergence coefficients, and to calculate the convergence coefficients by multiplying the common convergence coefficient by the respective coherence values. [Claim 2] The system according to claim 1, wherein the processing circuit configuration is configured to adapt the tap coefficients to estimate the transfer function between the speaker and one or more microphones. [Claim 3] The system according to claim 1, wherein the processing circuit configuration is configured to calculate the convergence coefficients based on the coherence value. [Claim 4] The system according to claim 1, wherein the processing circuit configuration is configured to evaluate the coherence value of a given tap based on a plurality of coefficient updates calculated for a given tap over a specified period of time. [Claim 5] The system according to any one of claims 1 to 4, comprising an eyeglass frame, wherein the microphone and the speaker are mounted at their respective positions on the eyeglass frame. [Claim 6] The system according to any one of claims 1 to 4, wherein the one or more microphones comprises a plurality of microphones, and the processing circuit configuration is configured to apply a beamforming function to the electrical signals output by the plurality of microphones in such a way as to emphasize selected sounds occurring within a selected angular range and suppress background sounds occurring outside the selected angular range. [Claim 7] A method for hearing assistance, The array of microphones (23, 24) that output electrical signals in response to acoustic waves incident on the microphones is attached near the head of the subject, The speaker (28) is attached near the ear of the subject, This includes amplifying and filtering the electrical signal to generate a drive signal for input to the speaker using a digital filter (100) having a plurality of taps, each having a tap coefficient selected to suppress feedback from the speaker to the microphone, estimating the coherence value of each tap coefficient over time, and adaptively calculating the tap coefficients while weighting the updates applied to the tap coefficients according to each coherence value, A method for calculating the tap coefficients, comprising applying the tap coefficients using a gradient descent method having respective convergence coefficients, and calculating the convergence coefficients, comprising multiplying a common convergence coefficient by the respective coherence values. [Claim 8] The method according to claim 7, wherein calculating the tap coefficients involves adapting the tap coefficients to estimate a transfer function between the speaker and one or more microphones (23, 24) in the array. [Claim 9] The method according to claim 7, comprising calculating the convergence coefficients based on the coherence values. [Claim 10] The method according to claim 7, comprising evaluating the coherence value of a given tap based on a plurality of coefficient updates calculated for a given tap over a specified period. [Claim 11] The method according to any one of claims 7 to 10, wherein the array of microphones (23, 24) and the speaker are mounted at their respective positions on the eyeglass frame. [Claim 12] The method according to any one of claims 7 to 10, wherein the array of microphones (23, 24) comprises a plurality of microphones, and includes applying a beamforming function to the electrical signals output by the plurality of microphones to enhance selected sounds occurring within a selected angular range and suppress background sounds occurring outside the selected angular range.

Images