Hearing aid comprising loop transfer function estimator and training method of loop transfer function estimator

Abstract

The present application discloses a hearing aid comprising a loop transfer function estimator and a training method of the loop transfer function estimator, the method comprising performing a plurality of training iterations, each training iteration comprising: obtaining simulation data from the hearing aid, the simulation data comprises at least one electrical input signal representing sound from a known, simulated acoustic environment of the hearing aid, a processed output signal indicating a frequency and / or level dependent gain function applied to the at least one electrical input signal, and a feedback path transfer function representing an impulse response of a feedback path of the hearing aid; determining a target open-loop transfer function according to the gain function changing along with the frequency and / or the level and the feedback path transfer function; determining a training open-loop transfer function according to at least one electrical input signal, the processed output signal, and a gain function that varies with frequency and / or level; and updating the ML prediction model according to the target open-loop transfer function and the training open-loop transfer function.

Description

Technical Field

[0001] The present application relates to the field of hearing aids and in particular to the estimation of a loop transfer function from the acoustic output to the input of a hearing aid. Background Art

[0002] In a hearing aid, the acoustic feedback problem creates an acoustic signal loop via the hearing aid forward path (from the input transducer (e.g., one or more microphones) to the output transducer (e.g., a speaker)) and the acoustic feedback path (from the output transducer to the input transducer).

[0003] The so-called open-loop transfer function describes important system characteristics, whose amplitude and phase variations over frequency are very relevant for controlling feedback in hearing aids.

[0004] Simple loop amplitude and / or phase estimates may be calculated as the difference between the signal amplitude / phase and these values ​​one loop delay earlier, as described in, for example, EP 3 291 581 A2.

[0005] However, this simple estimation is sensitive to the input signal going into the hearing aid input transducer (eg microphone). Certain input signals with amplitude / phase variations over time may lead to erroneously estimated open loop amplitude / phase values. Summary of the invention

[0006] In the present invention, a framework is proposed to estimate these open loop amplitude and phase values ​​using a machine learning (ML) based approach. The present invention is particularly concerned with estimating the loop transfer function from the acoustic output to the acoustic input of a hearing aid using machine learning techniques.

[0007] Hearing aids

[0008] A hearing aid is provided that includes a forward path for processing electrical signals representing sound.

[0009] The forward path comprises an input unit (IU) for receiving or providing at least one electrical input signal (y(n)) representing sounds in the hearing aid environment.

[0010] The forward path comprises a signal processing unit (PRO) configured to apply a frequency and / or level dependent gain ( g (n)). For example, n refers to time (eg a time index or a set of time indices). The signal processing unit (PRO) is configured to provide a processed output signal (u(n)) based thereon.

[0011] The forward path comprises an output transducer (OT) for generating a stimulus perceptible as sound by a user based on the processed output signal (u(n)).

[0012] The hearing aid further comprises an open loop transfer function estimator (OLTFE) comprising a trained ML prediction model configured to estimate an open loop transfer function (ξ′(ω,n)) based on at least one electrical input signal (y(n)) or a signal derived therefrom and a processed output signal (u(n)). For example, ω refers to a frequency (e.g., a frequency index or a set of frequency indices).

[0013] The prediction model is trained according to the methods disclosed herein.

[0014] The terms "according to" and "based on" are used interchangeably.

[0015] An improved hearing aid may thereby be provided.

[0016] The open loop transfer function can be interpreted as the transfer function of the signal propagating through the entire loop of the system. The gain ( g (n)) (e.g., a forward path gain function) may, for example, include gain contributions provided by one or more of: noise reduction, directionality (for multi-channel systems), different hearing loss compensation schemes, gain control algorithms, etc. In this specification, the term "gain" may mean amplification or attenuation (and may, for example, be implemented in the linear or logarithmic domain).

[0017] The input unit may comprise an input transducer such as a microphone and / or a wireless receiver.

[0018] In one or more exemplary hearing aids, at least one electrical input signal (y(n)) representing the sound of the hearing aid environment may be interpreted as a signal from a real sound environment, such as the sound environment that the hearing aid is located in. In other words, when the open loop transfer function estimator (OLTFE) includes the trained ML prediction model, the hearing aid may be operating in a normal operating mode.

[0019] The prediction model may be trained in a training mode of operation using simulation data from a known, simulated acoustic environment (e.g. situation). The training mode of operation may be initiated, for example, in the hearing aid via a user interface, or it may be performed during an offline period. In the present specification, simulation data may be interpreted as data generated as a result of a computer simulation using known inputs and known outputs (as opposed to data from a "real" acoustic situation or environment). This has the advantage that, for example, the contribution ( Figure 1A The v(n) in is known.

[0020] For example, a training operating mode (e.g., a training phase) may be followed by a normal operating mode (e.g., an inference phase). In other words, after the training phase, the weights of the prediction model may be fixed. In the normal operating mode (e.g., the inference phase), the prediction model is trained (e.g., the weights may be fixed) and ready for deployment. In the training operating mode, the weights of the ML prediction model may be updated based on simulation data.

[0021] In one or more exemplary hearing aids, an open-loop transfer function estimator (OLTFE) comprising a trained ML prediction model is configured to infer (e.g., deduce, estimate) an open-loop transfer function estimator (e.g., ξ′(ω,n)) based on at least one electrical input signal (y(n)) or a signal derived therefrom and a processed output signal (u(n)).

[0022] In one or more exemplary hearing aids, the hearing aid (e.g., an open-loop transfer function estimator) is configured to estimate an open-loop transfer function (ξ′(ω,n)) by applying a trained ML prediction model to at least one electrical input signal (y(n)) or a signal derived therefrom and a processed output signal (u(n)). The estimated open-loop transfer function (ξ′(ω,n)) may be regarded as an inferred output of the prediction model (e.g., an inferred ML output). The at least one electrical input signal (y(n)) or a signal derived therefrom and the processed output signal (u(n)) may be regarded as an inferred data set or data from a (real) acoustic environment. For example, the at least one electrical input signal (y(n)) or a signal derived therefrom and the processed output signal (u(n)) provided during a normal operating mode (e.g., an inference phase) is different from the at least one electrical input signal (y(n)) or a signal derived therefrom and the processed output signal (u(n)) included in the simulation data used to train the prediction model during a training operating mode (e.g., a training phase).

[0023] In one or more example hearing aids, the hearing aid (e.g., an open-loop transfer function estimator) is configured to determine an estimated open-loop transfer function (ξ′(ω,n)) by applying a trained ML prediction model to at least one electrical input signal (y(n)) or a signal derived therefrom and a processed output signal (u(n)).

[0024] In one or more exemplary hearing aids, the hearing aid further comprises a feedback control system configured to transfer the function ( h The feedback control system is configured to eliminate or reduce feedback in at least one electrical input signal (y(n)) via an acoustic or mechanical or electrical feedback path (FBP) from the output transducer to the input unit.

[0025] In one or more exemplary hearing aids, the feedback control system is configured to provide an estimate (v'(n)) of a current feedback signal (v(n)) received by an input unit via a feedback path (FBP). In one or more exemplary hearing aids, the feedback control system is configured to provide a feedback-corrected input signal (e(n)) based on at least one electrical input signal (y(n)) or a signal derived therefrom and an estimate (v'(n)) of the current feedback signal (v(n)). For example, the at least one electrical input signal (y(n)) can be written as x(n)+v(n), where v(n) is the current feedback signal received by the input unit via the feedback path (FBP). The hearing aid may not be aware of the feedback path transfer function. In one or more exemplary hearing aids, the feedback control system is configured to provide the feedback path transfer function ( h (n)) h '(n)).

[0026] In one or more exemplary hearing aids, the feedback path transfer function ( h (n)) represents the impulse response of a feedback path (FBP) from the output transducer (OT) to the input unit (IU) in at least one electrical input signal (y(n)).

[0027] In one or more exemplary hearing aids, the feedback control system includes an adaptive filter configured to provide a feedback path transfer function ( h (n)) (e.g., an estimate of the current feedback transfer function) h '(n)). For example, the adaptive filter may be configured to compensate for acoustic feedback from the output transducer (OT) to the input unit (IU).

[0028] In one or more exemplary hearing aids, an open-loop transfer function estimator (OLTFE) including a trained ML prediction model is configured to estimate an open-loop transfer function (ξ′(ω,n)) based on a feedback-corrected input signal (e(n)) and a processed output signal (u(n)). For example, the open-loop transfer function estimator (OLTFE) including a trained ML prediction model is configured to infer (e.g., deduce) an open-loop transfer function estimate (e.g., ξ′(ω,n)) based on the feedback-corrected input signal (e(n)) and the processed output signal (u(n)).

[0029] In one or more exemplary hearing aids, the hearing aid (e.g., an open-loop transfer function estimator) is configured to estimate an open-loop transfer function (ξ′(ω,n)) by applying a trained ML prediction model to a feedback-corrected input signal (e(n)) and a processed output signal (u(n)). The estimated open-loop transfer function (ξ′(ω,n)) may be viewed as an inferred output of the prediction model (e.g., an inferred ML output). The feedback-corrected input signal (e(n)) and the processed output signal (u(n)) may be viewed as inferred data or data from a “real” acoustic environment. For example, the feedback-corrected input signal (e(n)) and the processed output signal (u(n)) provided during a normal operating mode (e.g., an inference phase) are different from the feedback-corrected input signal (e(n)) and the processed output signal (u(n)) included in the simulation data used to train the prediction model during a training operating mode (e.g., a training phase).

[0030] In one or more exemplary hearing aids, the estimated open-loop transfer function includes an estimated open-loop amplitude (ξ′ M (ω,n)) and the estimated open-loop phase (ξ′ p (ω,n)).

[0031] In one or more exemplary hearing aids, the frequency and / or level dependent gain function (g(n)) is controlled based on the estimated open loop transfer function (ξ′(ω,n)). In other words, the frequency and / or level dependent gain function (g(n)) may be controlled based on the estimated open loop amplitude (ξ′ M (ω,n)) and the estimated open-loop phase (ξ′ p (ω,n)) for control.

[0032] In one or more exemplary hearing aids, the hearing aid is constituted by an air conduction hearing aid, a bone conduction hearing aid, or a combination thereof, or the hearing aid includes an air conduction hearing aid, a bone conduction hearing aid, or a combination thereof.

[0033] The hearing aid may be adapted to provide frequency dependent gain and / or level dependent compression and / or frequency transposition of one or more frequency ranges to one or more other frequency ranges (with or without frequency compression) to compensate for the hearing impairment of the user. The hearing aid may include a signal processor for enhancing an input signal and providing a processed output signal.

[0034] The hearing aid may comprise an output unit for providing a stimulus perceived by the user as an acoustic signal based on the processed electrical signal. The output unit may comprise an output transducer. The output transducer may comprise a receiver (speaker) for providing the stimulus as an acoustic signal to the user (e.g. in an acoustic (air conduction based) hearing aid). The output transducer may comprise a vibrator for providing the stimulus as a mechanical vibration of the skull to the user (e.g. in a bone attached or bone anchored hearing aid). The output unit may (in addition or as an alternative) comprise a (e.g. wireless) transmitter for transmitting the sound picked up by the hearing aid (e.g. via a network, e.g. in telephone operating mode) to another device, such as a remote communication partner.

[0035] The hearing aid may comprise an input unit for providing an electrical input signal representing a sound. The input unit may comprise an input transducer such as a microphone for converting the input sound into the electrical input signal. The input unit may comprise a wireless receiver for receiving a wireless signal comprising or representing a sound and providing the electrical input signal representing the sound.

[0036] The wireless receiver and / or transmitter may be configured to receive and / or transmit electromagnetic signals in the radio frequency range (3kHz to 300GHz), for example. The wireless receiver and / or transmitter may be configured to receive and / or transmit electromagnetic signals in the optical frequency range (e.g., infrared light 300GHz to 430THz or visible light such as 430THz to 770THz), for example.

[0037] A hearing aid may include a directional microphone system adapted to spatially filter sounds from the environment so as to enhance a target sound source among a plurality of sound sources in the local environment of a user wearing the hearing aid. The directional system may be adapted to detect (e.g., adaptively detect) from which direction a particular portion of the microphone signal originates. This may be implemented in a variety of different ways, such as described in the prior art. In hearing aids, microphone array beamformers are typically used to spatially attenuate background noise sources. The beamformer may include a linear constrained minimum variance (LCMV) beamformer. Many beamformer variants can be found in the literature. Minimum variance distortionless response (MVDR) beamformers are widely used in microphone array signal processing. Ideally, the MVDR beamformer keeps the signal from the target direction (also called the line of sight) unchanged, while maximally attenuating sound signals from other directions. The generalized sidelobe canceler (GSC) structure is an equivalent representation of the MVDR beamformer, which provides computational and digital representation advantages over direct implementation in its original form.

[0038] The hearing aid may comprise an antenna and a transceiver circuit which enables a wireless link to be established to an entertainment device (e.g. a television), a communication device (e.g. a telephone), a wireless microphone, a separate (external) processing device, or another hearing aid, etc. The hearing aid may thus be configured to wirelessly receive a direct electrical input signal from another device. Similarly, the hearing aid may be configured to wirelessly transmit a direct electrical output signal to another device. The direct electrical input or output signal may represent or comprise an audio signal and / or a control signal and / or an information signal.

[0039] In general, the wireless link established by the antenna and the transceiver circuit of the hearing aid may be of any type. The wireless link may be a link based on near field communication, such as an inductive link based on inductive coupling between antenna coils of a transmitter part and a receiver part. The wireless link may be based on far field electromagnetic radiation. Preferably, the frequency used to establish the communication link between the hearing aid and the other device is lower than 70 GHz, such as in the range from 50 MHz to 70 GHz, such as higher than 300 MHz, such as in the ISM range higher than 300 MHz, such as in the 900 MHz range or in the 2.4 GHz range or in the 5.8 GHz range or in the 60 GHz range (ISM = Industrial, Scientific and Medical, such standardized ranges are defined, for example, by the International Telecommunication Union ITU). The wireless link may be based on standardized or dedicated technologies. The wireless link may be based on Bluetooth technology (such as low power Bluetooth technology, such as LE Audio) or ultra-wideband (UWB) technology.

[0040] The hearing aid may consist of or form part of a portable (e.g. configured to be wearable) device, e.g. a device comprising a local energy source such as a battery, e.g. a rechargeable battery. The hearing aid may, e.g., be a low weight, easily wearable device, e.g., having a total weight of less than 100 g, e.g., less than 20 g, e.g., less than 5 g.

[0041] The hearing aid may include a "forward" (or "signal") path between the input and output of the hearing aid for processing audio signals. A signal processor may be located in the forward path. The signal processor may be adapted to provide a frequency-dependent gain according to the specific needs of the user (e.g. hearing loss). The hearing aid may include an "analysis" path having functional parts for analyzing signals and / or controlling processing of the forward path. Part or all of the signal processing in the analysis path and / or the forward path may be performed in the frequency domain, in which case the hearing aid includes appropriate analysis and synthesis filter banks. Part or all of the signal processing in the analysis path and / or the forward path may be performed in the time domain.

[0042] The analog electrical signal representing the acoustic signal can be converted into a digital audio signal in an analog-to-digital (AD) conversion process, where the analog signal is sampled at a predetermined sampling frequency or sampling rate f. s Sampling, f sFor example, in the range from 8 kHz to 48 kHz (adapted to the specific needs of the application) at discrete time points t n (or n) provides digital samples x n (or x[n]), each audio sample is passed through a predetermined N b The bit represents the sound signal at t n The value of N b For example, in the range from 1 to 48 bits, such as 24 bits. Each audio sample thus uses N b bit quantization (resulting in 2 Nb different possible values). A digital sample x has a 1 / f s The time length, such as 50μs, for f s = 20kHz. Multiple audio samples can be arranged in time frames. A time frame can include 64 or 128 audio data samples. Other frame lengths can be used depending on the actual application.

[0043] The hearing aid may include an analog-to-digital (AD) converter to digitize an analog input (e.g., from an input transducer such as a microphone) at a predetermined sampling rate, such as 20 kHz. The hearing aid may include a digital-to-analog (DA) converter to convert the digital signal into an analog output signal, such as for presentation to a user via an output transducer.

[0044] The hearing aid, such as the input unit and / or the antenna and transceiver circuitry, may comprise a transform unit for transforming a time domain signal into a signal in a transform domain (e.g. frequency domain or Laplace domain, Z transform, wavelet transform, etc.). The transform unit may consist of or include a time-frequency (TF) transform unit for providing a time-frequency representation of the input signal. The time-frequency representation may comprise an array or mapping of corresponding complex or real values ​​of the signal concerned in a specific time and frequency range. The TF transform unit may comprise a filter bank for filtering the (time-varying) input signal and providing a plurality of (time-varying) output signals, each output signal comprising a distinct frequency range of the input signal. The TF transform unit may comprise a Fourier transform unit (e.g. a discrete Fourier transform (DFT) algorithm, a short-time Fourier transform (STFT) algorithm, or a similar algorithm) for transforming the time-varying input signal into a (time-varying) signal in the (time-)frequency domain. The minimum frequency f is taken into account by the hearing aid. min To the maximum frequency f max The frequency range of may include a portion of the typical human hearing range from 20 Hz to 20 kHz, for example a portion of the range from 20 Hz to 12 kHz. Typically, the sampling rate f s Greater than or equal to the maximum frequency f max twice, that is, f s ≥2f maxThe signals of the forward path and / or analysis path of the hearing aid may be split into NI frequency bands (e.g. of uniform width), wherein NI is, for example, greater than 5, such as greater than 10, such as greater than 50, such as greater than 100, such as greater than 500, at least parts of which are processed separately. The hearing aid may be adapted to process the signals of the forward and / or analysis path in NP different frequency channels (NP≤NI). The frequency channels may be of uniform or non-uniform width (e.g. the width increases with frequency), overlapping or non-overlapping.

[0045] The hearing aid may be configured to operate in different modes, such as a normal mode and one or more specific modes, which may be selected by a user or may be automatically selected, for example. The operating mode may be optimized for a specific acoustic situation or environment, such as a communication mode, e.g., a telephone mode. The operating mode may include a low power mode, in which the functionality of the hearing aid is reduced (e.g., to save energy), such as disabling wireless communication and / or disabling specific features of the hearing aid.

[0046] The hearing aid may include a plurality of detectors configured to provide status signals related to the current network environment of the hearing aid (e.g., the current acoustic environment), and / or to the current state of a user wearing the hearing aid, and / or to the current state or operating mode of the hearing aid. Alternatively or additionally, one or more of the detectors may form part of an external device that communicates with the hearing aid (e.g., wirelessly). The external device may, for example, include another hearing aid, a remote control, an audio transmission device, a phone (e.g., a smartphone), an external sensor, etc.

[0047] One or more of the plurality of detectors may operate on a full-band signal (time domain). One or more of the plurality of detectors may operate on a band-split signal ((time-)frequency domain), for example in a limited number of frequency bands.

[0048] The plurality of detectors may include a level detector for estimating the current level of the signal of the forward path. The detector may be configured to determine whether the current level of the signal of the forward path is above or below a given (L-)threshold. The level detector acts on the full-band signal (time domain). The level detector acts on the band-split signal ((time-)frequency domain).

[0049] The hearing aid may comprise a voice activity detector (VAD) for estimating whether (or with what probability) an input signal (at a given point in time) comprises a voice signal. In the present specification, a voice signal may be taken to mean a signal comprising speech from a human being. It may also include other forms of vocalizations (such as singing) produced by a human speech system. The voice activity detector unit may be adapted to classify the user's current acoustic environment as a "voice" or a "no-voice" environment. This has the following advantage: time periods comprising an electric microphone signal of human vocalizations (such as speech) in the user's environment may be identified and thus separated from time periods comprising only (or mainly) other sound sources (such as artificially produced noise). The voice activity detector may be adapted to detect the user's own voice as "voice" as well. As an alternative, the voice activity detector may be adapted to exclude the user's own voice from the detection of "voice".

[0050] The hearing aid may include a self-voice detector for estimating whether (or with what probability) a particular input sound (e.g. voice, such as speech) originates from the voice of a user of the hearing system. The microphone system of the hearing aid may be adapted to be able to distinguish the user's own voice from the voice of another person and possibly from unvoiced sounds.

[0051] The plurality of detectors may include a motion detector, such as an accelerometer. The motion detector may be configured to detect movement of the user's facial muscles and / or bones, such as due to speech or chewing (eg, jaw movement), and provide a detector signal indicative of the movement.

[0052] The hearing aid may comprise a classification unit configured to classify the current situation based on an input signal from (at least part of) the detector and possibly other inputs. In this specification, a "current situation" may be defined by one or more of the following:

[0053] a) the physical environment (e.g. including the current electromagnetic environment, such as the presence of electromagnetic signals (including audio and / or control signals) intended or not intended to be received by the hearing aid, or other properties of the current environment other than acoustics);

[0054] b) Current acoustic conditions (input level, feedback, etc.);

[0055] c) the user’s current mode or state (motion, temperature, cognitive load, etc.);

[0056] d) The current mode or status of the hearing aid and / or another device communicating with the hearing aid (selected program, time elapsed since last user interaction, etc.).

[0057] The classification unit may be based on or may comprise a neural network, such as a recurrent neural network, such as a trained neural network.

[0058] Hearing aids may include acoustic (and / or mechanical) feedback control (e.g., suppression) or an echo cancellation system. Adaptive feedback cancellation has the ability to track changes in the feedback path over time. It is usually based on a linear time-invariant filter to estimate the feedback path, but the filter weights are updated over time. The filter updates can be calculated using a stochastic gradient algorithm, including some form of the least mean square (LMS) or normalized LMS (NLMS) algorithm. They all have the property of minimizing the error signal in terms of mean square, with NLMS additionally normalizing the filter updates to the square of the Euclidean norm of a reference signal.

[0059] The hearing aid may also include other appropriate functions for the application in question, such as compression, noise reduction, etc.

[0060] A hearing aid may comprise a hearing instrument, e.g. adapted to be located at the ear of a user or fully or partially in the ear canal. A hearing system may comprise a loudspeaker amplifier (comprising a plurality of input transducers (e.g. a microphone array) and a plurality of output transducers such as one or more loudspeakers, and one or more audio (and possibly video) transmitters, e.g. for use in an audio conferencing situation), e.g. comprising a beamformer filter unit, e.g. providing a plurality of beamforming capabilities.

[0061] application

[0062] In one aspect, there is provided a use of a hearing aid as described above, in detail in the "Detailed Description" section and in the claims. There may be provided an application in a system comprising one or more hearing aids (such as hearing instruments), a classroom amplification system, etc. There is also provided a use of a hearing aid in an application susceptible to acoustic feedback.

[0063] method

[0064] A training method for an ML prediction model for use in an open-loop transfer function estimator of a hearing aid (HD) is provided. The open-loop transfer function estimator (OLFTE) comprises an ML prediction model.

[0065] The method includes performing a plurality of training iterations.

[0066] Each training iteration of the plurality of training iterations includes obtaining simulated data from a hearing aid.

[0067] The simulation data includes at least one electrical input signal (y(n)), a processed output signal (u(n)), and a feedback path transfer function ( h(n)). The at least one electrical input signal (y(n)) represents a sound from a known, simulated acoustic environment of the hearing aid (HD). The at least one electrical input signal (y(n)) may be regarded as a known electrical input signal (e.g. known input data). The processed output signal (u(n)) represents a gain function (g(n)) that is frequency and / or level dependent and has been applied to the at least one electrical input signal (y(n)) or a signal derived therefrom. The feedback path transfer function ( h (n)) represents the impulse response of the feedback path (FBP) of the hearing aid. The feedback path transfer function ( h (n)) can be considered as a known feedback path transfer function (eg, verified only against simulation data).

[0068] Each of the multiple training iterations includes a training loop based on a frequency and / or level dependent gain function (g(n)) and a feedback path transfer function ( h (n)) Determine the target open-loop transfer function

[0069] Each training iteration of the plurality of training iterations comprises determining a training open-loop transfer function based on at least one electrical input signal (y(n)) or a signal derived therefrom, a processed output signal (u(n)), and a gain function (g(n)) that varies with frequency and / or level

[0070] The ML prediction model is configured to receive as input at least one electrical input signal (y(n)) or a signal derived therefrom, a processed output signal (u(n)), and a gain function (g(n)) that varies with frequency and / or level (e.g., part of the simulation data) and provide as output a trained open-loop transfer function.

[0071] Each of the multiple training iterations includes a training step based on the target open-loop transfer function and training open-loop transfer function Update ML prediction models.

[0072] In one or more example approaches, the ML prediction model can be viewed as a machine learning (ML) algorithm. For example, the ML prediction model can be viewed as a learning algorithm that includes a prediction model.

[0073] In one or more exemplary methods, the simulation data includes the applied frequency- and / or level-dependent gain function. In other words, the applied frequency- and / or level-dependent gain function can be inferred (e.g., determined) from the processed signal. For example, obtaining the simulation data includes determining the applied frequency- and / or level-dependent gain function based on the processed signal.

[0074] In one or more exemplary methods, the simulation data may include a current feedback signal (v(n)) received from the input unit (IU) via the feedback path (FBP). For example, the current feedback signal indicates a contribution from the feedback path (FBP) to the electrical input signal (y(n)). For example, the current feedback signal may indicate a feedback path transfer function ( h (n)).

[0075] In one or more exemplary methods, the simulation data includes one or more of the following: at least one electrical input signal (y(n)), a processed signal (u(n)), an applied frequency and / or level dependent gain function (g(n)), and a feedback path transfer function ( h (n)).

[0076] In one or more example methods, a target open-loop transfer function is determined based on known data, such as data from a known, simulated acoustic environment of a hearing aid. In other words, the target open-loop transfer function may be considered a required (e.g., desired) open-loop transfer function. For example, the target open-loop transfer function may be determined based on simulation data (e.g., a portion of the simulation data). In one or more example methods, a training open-loop transfer function may be determined based on simulation data (e.g., a portion of the simulation data). In one or more example methods, a prediction model may be trained using the target open-loop transfer function and the training open-loop transfer function.

[0077] In this specification, simulated data may refer to data generated as a result of a computer simulation using known inputs and known outputs (as opposed to data from a "real" acoustic situation or environment). This has the advantage that, for example, the contribution from the feedback path (FBP) to the input signal (y(n)) picked up by the input transducer (and mixed with the external signal (x(n))) is not Figure 1A For example, at least one electrical input signal may be written as x(n)+v(n), where v(n) is the current feedback signal received by the input unit via the feedback path (FBP). The current feedback signal may be known. The feedback path transfer function may be known.

[0078] In one or more exemplary methods, determining the target open-loop transfer function includes determining an impulse response (e.g., G(ω,n)) of the applied frequency and / or level-dependent gain function. In one or more exemplary methods, determining the target open-loop transfer function includes determining a feedback path transfer function ( h (n)) (e.g., H(ω,n)).

[0079] For example, the applied frequency- and / or level-dependent gain function may be viewed as a frequency- and / or level-dependent gain function applied in the frequency domain. In other words, the applied frequency- and / or level-dependent gain function may be in the time domain. For example, the frequency response of the feedback path transfer function may be viewed as a feedback path transfer function in the frequency domain. In other words, the feedback path transfer function may be in the time domain.

[0080] In one or more exemplary methods, a target open-loop transfer function (eg, ) includes determining the target open-loop transfer function as

[0081]

[0082] Where n refers to time, ω refers to frequency, and (·) refers to product (eg, multiplication).

[0083] For example, the target open-loop transfer function may be determined according to equation (1).

[0084] In one or more exemplary methods, the simulation data also includes a feedback-corrected input signal (e(n)) and a feedback path transfer function ( h (n)) h In one or more exemplary embodiments, the feedback corrected input signal (e(n)) indicates a signal in which acoustic or mechanical or electrical feedback originating from the feedback path (FBP) is reduced or eliminated.

[0085] In one or more exemplary methods, each of the plurality of training iterations includes a training step based on a frequency and / or level dependent gain function (g(n)), a feedback path transfer function ( h (n)), and the feedback path transfer function ( h (n)) h '(n)) Determine the target open-loop transfer function

[0086] In one or more exemplary methods, each of the plurality of training iterations includes determining a training open-loop transfer function based on a feedback-corrected input signal (e(n)), a processed output signal (u(n)), and a frequency and / or level-dependent gain function (g(n)).

[0087] For example, the method comprises obtaining a feedback-corrected input signal and an estimate of a feedback path transfer function from a feedback control system of the hearing aid. The feedback control system of the hearing aid may be configured to obtain an estimate of a feedback path transfer function ( h(n)) eliminates or reduces feedback in the at least one electrical input signal (y(n)). The feedback control system of the hearing aid may be configured to provide an estimate (v'(n)) of a current feedback signal (v(n)) received from the input unit (IU) via a feedback path (FBP). The feedback control system of the hearing aid may be configured to provide a feedback-corrected input signal (e(n)) based on (e.g. based on and / or as a function of) at least one electrical input signal (y(n)) or a signal derived therefrom and the estimate (v'(n)) of the current feedback signal. The feedback control system may be configured to provide a feedback path impulse response ( h (n)) h '(n)). For example, the feedback control system includes an adaptive filter (ALG, FIL, h '(n)), configured to provide a feedback path transfer function ( h (n)) h '(n)). In other words, the method comprises selecting from an adaptive filter (ALG, FIL, h '(n)) obtain an estimate of the feedback path transfer function.

[0088] In one or more exemplary methods, a target open-loop transfer function is determined The method includes determining a frequency response G(ω,n) of an applied gain function (g(n)) that varies with frequency and / or level. In one or more exemplary methods, determining a target open-loop transfer function This includes determining the feedback path transfer function ( h In one or more exemplary methods, the target open-loop transfer function is determined by This includes determining the feedback path transfer function ( h (n)) h '(n)) is the frequency response H′(ω,n).

[0089] For example, the frequency response of the applied frequency- and / or level-dependent gain function may be viewed as a frequency- and / or level-dependent gain function applied in the frequency domain. In other words, the applied frequency- and / or level-dependent gain function may be in the time domain. For example, the frequency response of the estimator of the feedback path transfer function may be viewed as an estimator of the feedback path transfer function in the frequency domain. In other words, the estimator of the feedback path transfer function may be in the time domain.

[0090] In one or more exemplary methods, a target open-loop transfer function is determined Including the target open-loop transfer function Determined as:

[0091]

[0092] Where n refers to time, ω refers to frequency, and (·) refers to product (eg, multiplication).

[0093] For example, the target open-loop transfer function may be determined according to equation (2).

[0094] In one or more exemplary methods, the ML prediction model includes a deep neural network (DNN). For example, the DNN may include at least two neural networks (e.g., layers). For example, the DNN may include one or more of the following: a convolutional neural network (CNN), a recurrent neural network (RNN). For example, the DNN may include one or more of the following: a convolution-based neural network, a recurrent-based neural network. The RNN may include a gated recurrent unit (GRU). In one or more exemplary methods, the ML prediction model includes one or more of the following: a DNN, a CNN, a RNN, and any other suitable neural network.

[0095] In one or more exemplary methods, a training open-loop transfer function is determined It includes providing at least one electrical input signal (y(n)) or a signal derived therefrom, a processed output signal (u(n)), and a gain function (g(n)) that varies with frequency and / or level as input to an ML prediction model.

[0096] In one or more exemplary methods, each training iteration includes applying at least one electrical input signal (y(n)) or a signal derived therefrom, a processed output signal (u(n)), and a frequency and / or level dependent gain function (g(n)) as inputs to an ML prediction model (e.g., ML model) to provide a trained open-loop transfer function (e.g., ML output).

[0097] In one or more exemplary methods, updating the ML prediction model includes updating the ML prediction model according to a target open-loop transfer function and training open-loop transfer function A training error signal is determined. In one or more example methods, updating the ML prediction model includes updating weights of the ML prediction model using a learning rule based on the training error signal.

[0098] In one or more exemplary methods, the method may include defining a loss function (e.g., a cost function) based on a target open-loop transfer function and a training open-loop transfer function to provide a training error signal. For example, the loss function of the ML prediction model may quantify the difference between the training open-loop transfer function (e.g., predicted by the ML prediction model) and the target open-loop transfer function (e.g., the expected output of the ML prediction model). In one or more examples, the training error signal indicates a training loss associated with the ML prediction model. Minimization of such a training loss (e.g., reducing the training error signal) may indicate that the prediction of the training open-loop transfer function is appropriate (e.g., satisfactory, meets the needs). In one or more exemplary methods, the loss function may be one or more of the following: mean square error (MSE), binary cross entropy (BCE) loss function, and any other suitable loss function.

[0099] The training open-loop transfer function can converge to the target open-loop transfer function by performing each training iteration of a plurality of training iterations of the method of the present invention.

[0100] In one or more example methods, training the prediction model includes updating weights of the ML prediction model based on the training error signal. For example, the weights of the ML prediction model may be updated (e.g., adjusted) when the training loss is minimized. For example, the weights of the ML prediction model may not be updated (e.g., adjusted) when the training loss is not minimized. The updated (e.g., adjusted) weights may be stored in a memory associated with the ML prediction model (e.g., a memory included in the ML prediction model (e.g., Figure 7 ML-PM module in )).

[0101] In one or more exemplary methods, the target open-loop transfer function can be interpreted as a reference open-loop transfer function or a true open-loop transfer function. For example, the target open-loop transfer function can be interpreted as a true open-loop transfer function or a reference open-loop transfer function in terms of comparing the target open-loop transfer function to the training open-loop transfer function. The target open-loop transfer function, the true open-loop transfer function, and the reference open-loop transfer function can be used interchangeably.

[0102] For example, the training open-loop transfer function may be referred to as the current open-loop transfer function in this application. The training open-loop transfer function and the current open-loop transfer function may be used interchangeably.

[0103] In one or more exemplary methods, the open-loop transfer function is trained Including training open loop amplitude and training open loop phase

[0104] In one or more exemplary methods, the target open-loop transfer function Including target open loop amplitude and the target open-loop phase For example, the target open-loop amplitude and the target open-loop phase may be determined (eg, calculated) according to equation (1). For example, the target open-loop amplitude and the target open-loop phase may be determined (eg, calculated) according to equation (2).

[0105] Optionally, updating the ML prediction model includes determining a training error signal based on the target open-loop amplitude, the target open-loop phase, the training open-loop amplitude, and the training open-loop phase.

[0106] For example, the loss function of the ML prediction model may quantify a first difference between the target open-loop amplitude and the training open-loop amplitude. For example, the loss function of the ML prediction model may quantify a second difference between the target open-loop phase and the training open-loop phase. For example, the weights of the ML prediction model may be updated (e.g., adjusted) when the first difference and the second difference are minimized.

[0107] For example, the target open-loop amplitude and the target open-loop phase It can be expressed in real and imaginary form. For example, the training open loop amplitude and training open loop phase Can be expressed in real-valued and imaginary-valued form.

[0108] In one or more exemplary methods, the method of the present invention is performed by an external device (eg, a computer). For example, the training method can be a computer-implemented method.

[0109] In one or more exemplary methods, the training of the ML prediction model can be performed during an offline training period. In other words, the known, simulated acoustic environment can be interpreted as a hearing aid environment that models (e.g., simulates) real-world conditions (e.g., situations). For example, the aforementioned known, simulated acoustic environment can be generated by computer simulation. For example, during an offline training period, it can be interpreted as a representative modeling of real-world conditions (e.g., situations) in a computer simulation.

[0110] In one or more exemplary methods, the method of the present invention may be performed using simulation data from a plurality of known, simulated acoustic environments (e.g., known, simulated acoustic situations). For example, the simulation data may be provided by a computer simulation of a hearing aid in an acoustic environment (e.g., or a plurality of acoustic environments). The simulation data may be interpreted as training data, for example, for training an ML prediction model.

[0111] In one or more exemplary methods, the simulation data may include data from multiple computer simulations (e.g., multiple known, simulated acoustic environments and / or multiple hearing aid systems). For example, the data from the multiple computer simulations may include different feedback path transfer functions ( h (n)), different frequency and / or level dependent gain functions ( g(n)), and different external input signals (x(n)) to the hearing aid, where the external input signal is the part of the electrical input signal that is not caused by feedback. For example, a gain function that varies with frequency and / or level can be viewed as a forward path processing gain function.

[0112] In one or more exemplary methods, the simulation data may include data from multiple sound sources. The multiple sound sources may include one or more of the following: noise sound as the incoming sound x(n) of the hearing aid, speech sound, music sound, sound recorded from daily life. The multiple sound sources may include one or more of the following: speech, noise, speech mixed with different types of noise in different amounts (e.g., to provide different signal-to-noise ratios (SNRs)), sounds from daily life (e.g., from different environments including multiple sound sources with multiple different mixtures).

[0113] In one or more exemplary methods, for a hearing aid without a feedback cancellation system, a target open-loop transfer function (e.g., including a target open-loop amplitude and a target open-loop phase) may be determined based on equation (1). In one or more exemplary methods, for a hearing aid including a feedback cancellation system, a target open-loop transfer function (e.g., including a target open-loop amplitude and a target open-loop phase) may be determined based on equation (2). For example, for a hearing aid without a feedback cancellation system and a hearing aid including a feedback cancellation system, the target open-loop transfer function (e.g., including a target open-loop amplitude and a target open-loop phase) may be calculated in a simulation setting. In addition, different hearing aid signals (e.g., y(n), u(n), e(n)) and the calculated amplitude and phase of the open-loop transfer function may be provided to the ML prediction model.

[0114] In other words, the simulation data can have many different sources (speech, noise, daily life, input, output, feedback compensation, etc.).

[0115] The trained ML prediction model may be the (final) result of training with simulation data. In other words, the ML prediction model may be trained by updating the ML prediction model during a training operating mode. For example, the training method may be performed during a training operating mode of a hearing aid.

[0116] The process of training with simulated data is Figure 2A The hearing aid with the resulting, e.g., trained, ML prediction model is exposed to a real signal from a real acoustic situation (e.g., environment) in Figure 2B Shown in.

[0117] The entire process can also be configured to train any function / value that varies with the open loop transfer function.

[0118] Computer readable medium or data carrier

[0119] The present invention further provides a tangible computer-readable medium (data carrier) storing a computer program including program code (instructions), which, when executed on a data processing system (computer), enables the data processing system to execute (implement) at least part (such as most or all) of the steps of the method described above, described in detail in the "Specific Implementation Method" and defined in the claims.

[0120] As an example but not limitation, the aforementioned tangible computer readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage device, or any other medium that can be used to execute or store the desired program code in the form of instructions or data structures and can be accessed by a computer. As used herein, disks include compact disks (CDs), laser disks, optical disks, digital versatile disks (DVDs), floppy disks, and blue-ray disks, wherein these disks usually reproduce data magnetically, while these disks can reproduce data optically with lasers. Other storage media include storage in DNA (e.g., in synthetic DNA chains). Combinations of the above disks should also be included in the scope of computer readable media. In addition to being stored on tangible media, computer programs can also be transmitted via transmission media such as wired or wireless links or networks such as the Internet and loaded into a data processing system to run at a location different from the tangible media.

[0121] Computer Programs

[0122] In addition, the present application provides a computer program (product) comprising instructions, which, when executed by a computer, causes the computer to execute (the steps of) the method described above, described in detail in the “Detailed Description” and defined in the claims.

[0123] Data processing system

[0124] On the one hand, the present invention further provides a data processing system comprising a processor and a program code, wherein the program code enables the processor to perform at least part (such as most or all) of the steps of the method described above, described in detail in the "Specific Implementation Method" and defined in the claims.

[0125] Hearing System

[0126] In another aspect, there is provided a hearing aid comprising the above description, the detailed description in the "Detailed Description of the Invention" and the claims, and a hearing system comprising an auxiliary device.

[0127] The hearing system may be adapted to establish a communication link between the hearing aid and the auxiliary device so that information (eg control and status signals, possibly audio signals) can be exchanged or forwarded from one device to the other.

[0128] The auxiliary device may include or may consist of a remote control, a smart phone, or other portable or wearable electronic device such as a smart watch or the like.

[0129] The auxiliary device may consist of or comprise a remote control for controlling the functions and operation of the hearing aid. The functions of the remote control are implemented in a smartphone, which may run an APP enabling control of the functions of the audio processing device via the smartphone (the hearing aid comprises a suitable wireless interface to the smartphone, e.g. based on Bluetooth or some other standardized or proprietary solution).

[0130] The auxiliary device may be constituted by or include an audio gateway device, which is suitable for receiving multiple audio signals (e.g. from an entertainment device such as a TV or music player, from a telephone device such as a mobile phone, or from a computer such as a PC, a wireless microphone, etc.) and is suitable for selecting and / or combining appropriate signals (or signal combinations) from the received audio signals for transmission to the hearing aid.

[0131] The auxiliary device may consist of or may comprise a further hearing aid.The hearing system may comprise two hearing aids adapted to implement a binaural hearing system, eg a binaural hearing aid system.

[0132] APP

[0133] On the other hand, the present invention also provides a non-transient application called APP. The APP includes executable instructions configured to run on an auxiliary device to implement a user interface for the hearing aid or hearing system described above, described in detail in the "Detailed Description of the Invention" and defined in the claims. The APP can be configured to run on a mobile phone such as a smart phone or another portable device that enables communication with the hearing aid or hearing system.

[0134] definition

[0135] In this specification, a hearing aid, such as a hearing instrument, refers to a device suitable for improving, enhancing and / or protecting the hearing ability of a user by receiving an acoustic signal from the user's environment, generating a corresponding audio signal, possibly modifying the audio signal, and providing the possibly modified audio signal as an audible signal to at least one ear of the user. The audible signal may be provided, for example, in the form of an acoustic signal radiated into the user's outer ear, an acoustic signal transmitted to the user's inner ear as mechanical vibrations through the bone structure of the user's head and / or through parts of the middle ear, and an electrical signal transmitted directly or indirectly to the user's cochlear nerve.

[0136] The hearing aid may be configured to be worn in any known manner, such as as a unit worn behind the ear (with a tube directing the radiated acoustic signal into the ear canal or with an output transducer such as a loudspeaker arranged close to or in the ear canal), as a unit arranged wholly or partly in the auricle and / or ear canal, as a unit connected to a fixed structure implanted in the skull, such as a vibrator, or as a connectable or wholly or partly implanted unit, etc. The hearing aid may comprise a single unit or several units that communicate with each other (e.g. acoustically, electrically or optically). The loudspeaker may be arranged in a housing together with the other components of the hearing aid, or it may itself be an external unit (possibly in combination with a flexible guiding element such as a dome-shaped element).

[0137] The hearing aid may be adapted to the needs of a specific user, such as hearing loss. The configurable signal processing circuit of the hearing aid may be adapted to apply frequency- and level-dependent compression amplification of the input signal. The customized frequency- and level-dependent gain (amplification or compression) may be determined during the fitting process by the fitting system based on the user's hearing data, such as an audiogram, using the basic principles of fitting (e.g. adaptation to speech). The frequency- and level-dependent gain may, for example, be embodied in a processing parameter, uploaded to the hearing aid, for example, via an interface to a programming device (fitting system), and used by a processing algorithm executed by the configurable signal processing circuit of the hearing aid.

[0138] A "hearing system" refers to a system comprising one or two hearing aids. A "binaural hearing system" refers to a system comprising two hearing aids and adapted to provide audible signals to both ears of a user in a coordinated manner. A hearing system or binaural hearing system may also include one or more "auxiliary devices" that communicate with the hearing aids and affect and / or benefit from the functions of the hearing aids. The aforementioned auxiliary devices may include at least one of the following: a remote control, a remote microphone, an audio gateway device, an entertainment device such as a music player, a wireless communication device such as a mobile phone (e.g., a smart phone) or a tablet computer or another device, for example including a graphical interface. A hearing aid, a hearing system or a binaural hearing system may be used, for example, to compensate for the loss of hearing ability of a hearing-impaired person, to enhance or protect the hearing ability of a normal hearing person, and / or to transmit electronic audio signals to a person. A hearing aid or a hearing system may, for example, form part of or interact with a broadcast system, a classroom amplification system, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0139] Various aspects of the present invention will be best understood from the detailed description below in conjunction with the accompanying drawings. For clarity, the drawings are schematic and simplified, and only the details necessary for understanding the present invention are given, while other details are omitted. Throughout the specification, the same reference numerals are used for the same or corresponding parts. The various features of each aspect may be combined with any or all features of the other aspects. These and other aspects, features and / or technical effects will be apparent from and illustrated in conjunction with the following figures, in which:

[0140] Figure 1A A closed loop hearing aid system without a feedback cancellation system is shown;

[0141] Figure 1B A single channel hearing aid including a feedback cancellation system is shown;

[0142] Figure 2A A machine learning framework for training and predicting the open-loop magnitude and phase of an open-loop transfer function estimator is shown;

[0143] Figure 2B Shown including according to Figure 2A Trained open-loop transfer function estimator for hearing aids;

[0144] Figure 3 An example of a multi-channel hearing system is shown;

[0145] Figure 4 Schematically shows a RITE-type embodiment of a hearing aid according to the present invention;

[0146] Figure 5A A flow chart showing an exemplary method for estimating an open loop transfer function performed by a hearing aid without a feedback cancellation system according to the present invention;

[0147] Figure 5B A flow chart showing an exemplary method for estimating an open loop transfer function performed by a hearing aid including a feedback cancellation system according to the present invention;

[0148] Figure 5C A flow chart showing an exemplary method for estimating an open-loop transfer function performed by a hearing aid comprising a multi-channel hearing aid system according to the present invention;

[0149] Figures 6A-6C A flowchart showing an exemplary training method of an ML prediction model for use in an open-loop transfer function estimator of a hearing aid according to the present invention;

[0150] Figure 7 An exemplary structure of an ML prediction model according to the present invention is schematically shown.

[0151] By the detailed description given below, the further scope of application of the present invention will be apparent. However, it should be understood that while the detailed description and specific examples show the preferred embodiments of the present invention, they are only provided for illustrative purposes. For those skilled in the art, based on the following detailed description, other embodiments of the present invention will be apparent. DETAILED DESCRIPTION

[0152] The detailed description proposed below in conjunction with the accompanying drawings serves as a description of a variety of different configurations. The detailed description includes specific details for providing a thorough understanding of a number of different concepts. However, it is apparent to those skilled in the art that these concepts can be implemented without these specific details. Several aspects of the apparatus and method are described by a number of different blocks, functional units, modules, components, circuits, steps, processes, algorithms, etc. (collectively referred to as "elements"). Depending on the specific application, design limitations or other reasons, these elements can be implemented using electronic hardware, computer programs or any combination thereof.

[0153] The electronic hardware may include microelectromechanical systems (MEMS), (e.g., application specific) integrated circuits, microprocessors, microcontrollers, digital signal processors (DSPs), field programmable gate arrays (FPGAs), programmable logic devices (PLDs), gating logic, discrete hardware circuits, printed circuit boards (PCBs) (e.g., flexible PCBs), and other suitable hardware configured to perform a number of different functions described in this specification, such as sensors for sensing and / or recording physical properties of the environment, device, user, etc. Computer programs shall be broadly construed as instructions, instruction sets, codes, code segments, program codes, programs, subroutines, software modules, applications, software applications, software packages, routines, subroutines, objects, executables, execution threads, programs, functions, etc., whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.

[0154] The present application relates to the field of hearing aids and in particular to the estimation of a loop transfer function from the acoustic output to the input of a hearing aid.

[0155] Figure 1A shows a closed loop hearing aid system without a feedback cancellation system, Figure 1B A single channel hearing aid including a feedback cancellation system is shown.

[0156] Figure 1A A simplified diagram of an exemplary hearing aid including a forward path and a feedback path FBP that together form a closed loop is shown. For example, the feedback path FBP may include a transfer function h(n). The forward path comprises an input transducer M, such as a microphone, for picking up sound from the environment of the hearing aid and providing an electrical input signal y(n) representing the sound. The forward path further comprises a processor PRO for applying a gain to the electrical input signal (or a signal derived therefrom) and providing a processed signal u(n) therefrom. For example, the processor PRO is configured to apply a gain function g(n) that varies with frequency and / or level to the electrical input signal (or a signal derived therefrom). The forward path further comprises an output transducer SPK, such as a loudspeaker, for providing a stimulus that can be perceived as sound by the user based on the processed signal u(n). The time-varying transfer function of the forward path (e.g., the applied gain function that varies with frequency and / or level) and the time-varying transfer function of the feedback path (e.g., the feedback path transfer function) are respectively denoted as g (n) and h (n), where n represents time. The applied frequency- and / or level-dependent gain function g(n) (e.g., the applied frequency- and / or level-dependent gain function that specifies the impulse response of the forward path, such as the gain impulse response) (referred to herein as the "gain function g (n)”) and a feedback path transfer function (e.g., specifying the impulse response of the feedback path) h (n) are vectors whose elements represent the time-dependent response to an external change (in this particular case, the response to an impulse). Each vector represents a sample of the impulse response at time n. In a dynamic system, the impulse response may vary with time (making the impulse response vector dependent on the time index n). In a static system, the impulse response will remain constant and the impulse response vector will be independent of the time index n.

[0157] Closed-loop hearing aid system without feedback cancellation system operating in training mode

[0158] Closed-loop hearing aid systems without feedback cancellation systems e.g. Figure 1A The hearing aid of the present invention may be considered as a hearing aid operating in a training operating mode. In other words, the simulation data may be provided by a computer simulation that simulates the hearing aid in a known, simulated acoustic environment (e.g., modeling a real-world condition or environment). When the hearing aid is operating in a normal operating mode, the ML prediction model used in the open-loop transfer function estimator OLTFE may be trained using the simulation data from the known, simulated acoustic environment. For example, in the training operating mode, the weights of the ML prediction model may be updated based on the simulation data.

[0159] For example, the simulation data includes an electrical input signal, a processed output signal, and a feedback path transfer function. For example, the electrical input signal y(n) represents sound from a known, simulated acoustic environment of the hearing aid. For example, the processed output signal u(n) indicates a frequency and / or level dependent gain function g(n) applied to the electrical input signal y(n). The simulation data may include the frequency and / or level dependent gain function g(n). Optionally, the frequency and / or level dependent gain function g(n) may be inferred from the electrical input signal y(n) and the processed output signal u(n). For example, the feedback path transfer function h (n) represents the impulse response of the feedback path FBP of the hearing aid.

[0160] Target open-loop transfer function ( Figure 7 of ) can be used to train the prediction model. For example, the target open-loop transfer function can be based on a gain function g(n) that varies with frequency and / or level and a feedback path transfer function h (n) Determine.

[0161] The target open-loop transfer function can be determined as:

[0162] ξ Targ (ω,n)=G(ω,n)·H(ω,n) (1)

[0163] Where G(ω,n) is the frequency response of the gain function g(n) at time n, which is a function of frequency and / or level, and H(ω,n) is the feedback path transfer function h (n) is the frequency response at time n, ω is the frequency, and (·) is the product operator.

[0164] A training open-loop transfer function may be used to train the prediction model. For example, the training open-loop transfer function may be determined based on the electrical input signal y(n), the processed output signal u(n), and a gain function g(n) that varies with frequency and / or level.

[0165] For example, the ML prediction model is configured to receive simulation data as input and provide a trained open-loop transfer function as output.

[0166] For example, Figure 1A The system may include an open-loop transfer function estimator OLTFE in a training operation mode, which includes a configuration to provide a Figure 7 The training open-loop transfer function (For example, training open loop amplitude and training open loop phase )’s ML prediction model.

[0167] Figure 7 The target open-loop transfer function is Especially its magnitude and Phase It can be calculated according to equation (2). For example, the target open-loop transfer function may also be referred to as a true open-loop transfer function or a reference open-loop transfer function.

[0168] Closed-loop hearing aid without feedback cancellation system operating in normal operation mode

[0169] Closed-loop hearing aid systems without feedback cancellation systems e.g. Figure 1A The hearing aid of the present invention can be regarded as a hearing aid operating in a normal operating mode. In other words, the open-loop transfer function estimator OLTFE of the hearing aid may include a trained ML prediction model (eg, the weights of the ML prediction model may be fixed) and is ready to be deployed in the hearing aid.

[0170] For example, Figure 1A systems (such as Figure 1A ) includes an open-loop transfer function estimator OLTFE, which includes a trained ML prediction model for estimating an open-loop transfer function ξ′(ω,n) based on test data from a real sound environment (e.g., a situation), such as the sound of a hearing aid environment. For example, the open-loop transfer function estimator OLTFE receives as input an electrical input signal y(n) and a processed signal u(n), and outputs an estimated open-loop transfer function ξ′(ω,n) (e.g., an estimated open-loop amplitude ξ′ M (ω,n) and the estimated open-loop phase ξ′ p (ω,n)) is provided as an output. An estimated open-loop transfer function ξ′(ω,n) (e.g., an estimated open-loop amplitude ξ′ M (ω,n) and the estimated open-loop phase ξ′ p (ω, n)) can be used to modify (e.g. adjust, update) the frequency- and / or level-dependent gain function g(n) of the signal processing unit PRO. If, for example, at certain frequencies ω, the open-loop amplitude approaches or exceeds 1 (0 dB) and / or the open-loop phase approaches 0 degrees, i.e., positive feedback may occur, the frequency- and / or level-dependent gain function g(n) can be changed so that its amplification at these frequencies ω will be reduced.

[0171] In other words, the frequency and / or level dependent gain function g(n) may be modified when the open loop amplitude approaches or exceeds 1 (0 dB) and / or the open loop phase approaches 0 degrees at certain frequencies ω. For example, the frequency and / or level dependent gain function g(n) may be modified based on an estimate that the open loop amplitude approaches or exceeds 1 (0 dB) and / or the open loop phase approaches 0 degrees at certain frequencies ω.

[0172] Optionally, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop amplitude at certain frequencies is approximately equal to or greater than an amplitude threshold (e.g., not necessarily 0 dB). For example, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop phase at certain frequencies approximately satisfies a phase threshold (e.g., not necessarily 0 degrees).

[0173] For example, the frequency and / or level dependent gain function g(n) may be modified when the estimated open loop amplitude is approximately equal to 0 dB. Alternatively, the frequency and / or level dependent gain function g(n) may be modified when the estimated open loop amplitude is within an amplitude range including a lower limit and an upper limit. The lower limit of the amplitude range may be approximately equal to -10 dB, -6 dB, -3 dB, or 0 dB. The upper limit of the amplitude range may be approximately equal to +3 dB, +6 dB, +10 dB, or a value greater than +10 dB. The amplitude range may include the following ranges: [-10dB, +3dB], [-10dB, +6dB], [-10dB, +10dB], [-10dB, > +10dB], [-6dB, +3dB], [-6dB, +6dB], [-6dB, +10dB], [-10dB, > +10dB], [-3dB, +3dB], [-3dB, +6dB], [-3dB, +10dB], [-3dB, > +10dB], [0dB, +3dB], [0dB, +6dB], [0dB, +10dB], [0dB, > +10dB]. For example, “> +10dB” may be considered a value greater than 10dB. For example, the amplitude range may include a range of [-20dB, 0dB].

[0174] For example, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop phase is approximately equal to 0 degrees. Alternatively, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop phase is within a phase range. For example, the phase range may include the following ranges: [-180 degrees, +180 degrees], [-30 degrees, +30 degrees], [-60 degrees, 60 degrees], [-60 degrees, 30 degrees], and [-30 degrees, 60 degrees].

[0175] For example, the modification of the frequency- and / or level-dependent gain function g(n) may be different with frequency (e.g. because g (n) is a function of frequency).

[0176] exist Figure 1A In the embodiment, ξ' M (ω,n) and ξ' P (ω,n) refer to the estimated open-loop amplitude ξ′ M (ω,n) and the estimated open-loop phase ξ′ p(ω,n).

[0177] Figure 1B and Figure 1A Similarly, a single channel hearing aid is shown, but Figure 1B The hearing aid further comprises a feedback cancellation system.

[0178] Hearing aid processing through the processing functions of the processor PRO g The acoustic feedback path FBP from the receiver (speaker SPK) to the microphone M can be expressed by the feedback path transfer function h (n) represents the feedback path transfer function h Estimation of (n) h The adaptive filter (ALG, FIL) of the feedback path transfer function '(n) (e.g., representing the impulse response of an estimate of the feedback path FBP) is applied to the feedback path transfer function h (n), a real and practically unknown feedback path FBP is modeled. For example, the adaptive filter comprises an adaptive algorithm ALG and a variable filter FIL whose filter coefficients are determined (repeatedly updated) by the adaptive algorithm ALG according to the error signal e(n) and a reference signal, here the processed signal u(n). For example, the error signal e(n) is the input signal of the feedback correction, for example the electrical input signal y(n) from the microphone M minus the feedback estimate v'(n) provided by the variable filter FIL according to the processed signal u(n). The signal processing unit PRO may comprise, for example, a control unit according to the invention Figure 2A The open loop transfer function estimator OLTFE may form part of the processor or may be a separate unit (eg Figure 2A (same as in ).

[0179] Single channel hearing aid including feedback cancellation system running in training mode

[0180] Single channel hearing aid systems including feedback cancellation systems such as Figure 1B The hearing aid of the present invention may be considered as a hearing aid operating in a training operating mode. In other words, the simulation data may be provided by a computer simulation that simulates the hearing aid in a known, simulated acoustic environment (e.g., modeling a real-world condition or environment). When the hearing aid is operating in a normal operating mode, the ML prediction model used in the open-loop transfer function estimator OLTFE may be trained using the simulation data from the known, simulated acoustic environment. For example, in the training operating mode, the weights of the ML prediction model may be updated based on the simulation data.

[0181] For example, the simulation data includes a feedback-corrected input signal, a processed output signal, a feedback path transfer function, and an estimate of the feedback path transfer function.

[0182] For example, the feedback-corrected input signal e(n) refers to a signal in which acoustic or mechanical or electrical feedback is reduced or eliminated, the acoustic or mechanical or electrical feedback originating from the feedback path. The electrical input signal y(n) (e.g., representing sound from a known, simulated acoustic environment of the hearing aid) may include the aforementioned acoustic or mechanical or electrical feedback originating from the feedback path, so that the feedback-corrected input signal is determined based on the electrical input signal. For example, the processed output signal u(n) indicates a frequency- and / or level-dependent gain function g(n) applied to the feedback-corrected input signal. The simulation data may include the frequency- and / or level-dependent gain function g(n). Optionally, the frequency- and / or level-dependent gain function g(n) may be inferred from the feedback-corrected input signal and the processed output signal. For example, the feedback path transfer function h (n) represents the impulse response of the feedback path FBP of the hearing aid. For example, the feedback path transfer function h Estimation of (n) h’ (n) represents an estimate of the impulse response of the feedback path FBP of the hearing aid.

[0183] Target open-loop transfer function ( Figure 7 of ) can be used to train the prediction model. For example, the target open-loop transfer function can be determined based on a gain function that varies with frequency and / or level, a feedback path transfer function, and an estimate of the feedback path transfer function.

[0184] The target open-loop transfer function can be determined as:

[0185] ξ Targ (ω,n)= G(ω,n)·(H(ω,n)- H'(ω,n)) (2)

[0186] Where G(ω,n) is the frequency response of the gain function g(n) at time n, which is a function of frequency and / or level, and H(ω,n) is the feedback path transfer function h (n) is the frequency response at time n, H'(ω,n) is the frequency response of the estimator of the feedback path transfer function at time n, ω is the frequency, and (·) is the product operator.

[0187] Figure 7 The target open-loop transfer function is Especially its magnitude and Phase It can be calculated according to equation (2).

[0188] A training open-loop transfer function may be used to train the prediction model. For example, the training open-loop transfer function may be determined based on the feedback-corrected input signal e(n), the processed output signal u(n), and a gain function g(n) that varies with frequency and / or level.

[0189] For example, the ML prediction model is configured to receive simulation data as input and provide a trained open-loop transfer function as output.

[0190] For example, Figure 1B The system may include an open-loop transfer function estimator OLTFE in a training operation mode, which includes a configuration to provide a Figure 7 The training open-loop transfer function (For example, training open loop amplitude and training open loop phase )’s ML prediction model.

[0191] Single channel hearing aid including feedback cancellation system operating in normal operating mode

[0192] Single channel hearing aid systems including feedback cancellation systems such as Figure 1B The hearing aid of the present invention can be regarded as a hearing aid operating in a normal operating mode. In other words, the open-loop transfer function estimator OLTFE of the hearing aid may include a trained ML prediction model (eg, the weights of the ML prediction model may be fixed) and is ready to be deployed in the hearing aid.

[0193] For example, the open-loop transfer function estimator OLTFE includes a trained ML prediction model that estimates the open-loop transfer function ξ′(ω,n) based on test data from a real acoustic environment (e.g., a situation), such as the sound of a hearing aid environment. For example, the open-loop transfer function estimator OLTFE receives as input the feedback-corrected input signal e(n) and the processed signal u(n), and outputs an estimated open-loop transfer function ξ′(ω,n) (e.g., an estimated open-loop amplitude ξ′ M (ω,n) and the estimated open-loop phase ξ′ p (ω, n)) is provided as an output. The open loop transfer function estimator OLTFE may be included in the signal processing unit PRO or may be in communication with the signal processing unit PRO (eg Figure 1A The estimated open-loop transfer function ξ′(ω,n) (e.g., the estimated open-loop amplitude ξ′) M (ω,n) and the estimated open-loop phase ξ′ p (ω, n)) can be used to modify (e.g. adjust, update) the frequency- and / or level-dependent gain function g(n) of the signal processing unit PRO. If, for example, at certain frequencies ω, the open-loop amplitude approaches or exceeds 1 (0 dB) and / or the open-loop phase approaches 0 degrees, i.e., positive feedback may occur, the frequency- and / or level-dependent gain function g(n) can be changed so that its amplification at these frequencies ω will be reduced.

[0194] In other words, the frequency and / or level dependent gain function g(n) may be modified when the open loop amplitude approaches or exceeds 1 (0 dB) and / or the open loop phase approaches 0 degrees at certain frequencies ω. For example, the frequency and / or level dependent gain function g(n) may be modified based on an estimate that the open loop amplitude approaches or exceeds 1 (0 dB) and / or the open loop phase approaches 0 degrees at certain frequencies ω.

[0195] Optionally, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop amplitude at certain frequencies is approximately equal to or greater than an amplitude threshold (e.g., not necessarily 0 dB). For example, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop phase at certain frequencies approximately satisfies a threshold phase (e.g., not necessarily 0 degrees).

[0196] For example, the frequency and / or level dependent gain function g(n) may be modified when the estimated open loop amplitude is approximately equal to 0 dB. Alternatively, the frequency and / or level dependent gain function g(n) may be modified when the estimated open loop amplitude is within an amplitude range including a lower limit and an upper limit. The lower limit of the amplitude range may be approximately equal to -10 dB, -6 dB, -3 dB, or 0 dB. The upper limit of the amplitude range may be approximately equal to +3 dB, +6 dB, +10 dB, or a value greater than +10 dB. The amplitude range may include the following ranges: [-10dB, +3dB], [-10dB, +6dB], [-10dB, +10dB], [-10dB, > +10dB], [-6dB, +3dB], [-6dB, +6dB], [-6dB, +10dB], [-10dB, > +10dB], [-3dB, +3dB], [-3dB, +6dB], [-3dB, +10dB], [-3dB, > +10dB], [0dB, +3dB], [0dB, +6dB], [0dB, +10dB], [0dB, > +10dB]. For example, “> +10dB” may be considered a value greater than 10dB. For example, the amplitude range may include a range of [-20dB, 0dB].

[0197] For example, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop phase is approximately equal to 0 degrees. Alternatively, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop phase is within a phase range. For example, the phase range may include the following ranges: [-180 degrees, +180 degrees], [-30 degrees, +30 degrees], [-60 degrees, 60 degrees], [-60 degrees, 30 degrees], and [-30 degrees, 60 degrees].

[0198] For example, the modification of the frequency- and / or level-dependent gain function g(n) may be different with frequency (e.g. becauseg (n) is a function of frequency).

[0199] For example, the open-loop transfer function ξ′(ω,n) (e.g., the open-loop amplitude ξ′ M (ω,n) and the open-loop phase ξ′ p (ω,n)) can be estimated using the method described in EP3291581A2, or they can be determined using a machine learning based method as described in this application (eg also below).

[0200] In a simulation setup and / or during a training run mode, the feedback path transfer function h (n) (eg the impulse response of the feedback path FBP of a hearing aid) is known, in contrast to the actual situation, eg the real acoustic environment (eg it is therefore possible to calculate ξ(ω,n) in a simulation, eg a computer simulation).

[0201] The idea is to create many simulation scenarios, for example with different feedback paths h (n), different frequency- and / or level-dependent gain functions g(n), and different hearing aid input signals x(n). More specifically, the frequency- and / or level-dependent gain function g(n) (e.g., forward path gain function) includes possible noise reduction, directionality (e.g., for multi-channel systems), hearing loss compensation schemes, and gain control algorithms, etc.

[0202] For all of these above mentioned simulation scenarios and their combinations, the adaptive filter operates in the usual way to compensate for the simulated acoustic feedback originating from the feedback path.

[0203] Machine learning training framework Figure 2A For example, Figure 2A After the machine learning training framework shown, Figure 7 An example structure of an ML prediction model is shown.

[0204] Figure 2A shows a machine learning framework for training and predicting the open-loop magnitude and phase of an open-loop transfer function estimator, Figure 2B Shown including according to Figure 2A A trained open-loop transfer function estimator (eg, a trained ML prediction model) for a hearing aid.

[0205] For example, the open-loop transfer function estimator includes an ML prediction model. In other words, the open-loop transfer function estimator may include an ML prediction model. Figure 2A A training run mode is shown. Figure 2B The normal operating mode is shown.

[0206] exist Figure 2AIn this paper, the training and test signals are e(n) or u(n).

[0207] The calculated open-loop transfer function (e.g., including the calculated open-loop amplitude and the calculated open-loop phase) can be regarded as a target open-loop transfer function (e.g., including the target open-loop amplitude and the target open-loop phase), for example, determined according to equations (1), (2) (e.g., or, (3)).

[0208] The training signal may be interpreted as a signal used to train the ML prediction model, for example, to determine a training open-loop transfer function. For example, the ML prediction model may be trained using the feedback-corrected input signal e(n) and the processed signal u(n) (e.g., in combination with Figure 1B Optionally, the ML prediction model can be trained using the electrical input signal y(n) and the processed output signal u(n) (e.g., combined with Figure 1A described), which is not Figure 2A It is clearly shown in , however, is a possible scenario.

[0209] For example, the training open-loop transfer function may be determined based on the electrical input signal y(n) and the gain function g(n) that varies with frequency and / or level. For example, the training open-loop transfer function may be determined based on the feedback-corrected input signal e(n) and the gain function g(n) that varies with frequency and / or level. For example, the training open-loop transfer function may be determined based on the processed output signal u(n) and the gain function g(n) that varies with frequency and / or level. For example, when the ML prediction model has a memory over time (e.g., the memory of the ML prediction model includes training signals used to train the ML prediction model in previous training iterations), the training open-loop transfer function may be determined based on the electrical input signal y(n) or the feedback-corrected input signal e(n) or the processed output signal u(n), e.g., without the gain function g(n) that varies with frequency and / or level. The ML prediction model may include a layer having a memory of previous training signals, thereby enabling the ML prediction model to be trained using the electrical input signal y(n), the processed output signal u(n), or the feedback-corrected input signal e(n). Therefore, any signal between the microphone and the receiver SPK can be used to train the ML prediction model (for Figure 1A-3 any hearing system).

[0210] The test signal may be interpreted as a signal used during a normal operating mode of the hearing aid, such as a signal to be applied to a trained ML prediction model to provide an estimate of the open-loop transfer function. For example, the trained ML prediction model may receive as input a feedback-corrected input signal e(n) and a processed signal u(n), such as a signal from a real acoustic environment (e.g., in combination with Figure 1B For example, a trained ML prediction model may receive as input an electrical input signal y(n) and a processed signal u(n) (e.g., in combination with Figure 1A described), which is not Figure 2A It is clearly shown in , however, is a possible scenario.

[0211] Based on the trained ML prediction model, the open-loop amplitude and phase estimators ξ' M (ω,n) and ξ' P (ω,n) is obtained using a signal from the real situation (e.g., the environment) such as the feedback-corrected signal e(n) or the processed signal u(n). For example, Figure 2B The open-loop amplitude and phase estimator ξ' M (ω,n) and ξ' P (ω,n) and Figure 1A The estimated open-loop amplitude ξ M (ω,n) and the estimated open-loop phase ξ p (ω,n) is similar. For example, Figure 2A The open-loop amplitude and phase estimates ξ^ are calculated Targ,M (ω,n) and ξ^ Targ,P (ω,n) is similar to Figure 7 Target open loop amplitude and the target open-loop phase

[0212] Furthermore, the same training and prediction framework can also be used for multi-channel hearing aids, with the only difference being the calculation (eg, estimation) of the open-loop transfer function ξ(ω,n).

[0213] Figure 2B Shown including according to Figure 2A A hearing aid HD with a trained open-loop transfer function estimator, wherein the ML prediction model of the hearing aid is an ML prediction model trained based on simulated data from a known, simulated acoustic environment. The trained ML prediction model forms an open-loop transfer function estimator (see Figure 1A The OLTFE in FIG. 1 is a part of the OLTFE in FIG. 1 , which is used to calculate the gain function of the hearing aid according to the (real, non-simulated) signals such as e(n) and u(n) from the real situation (and the gain function of the current application g (n)), or y(n) and u(n) from the real situation (and the gain function of the current application g (n)), or any of the signals e(n), u(n), and y(n) from the real situation (and the currently applied gain function g (n)) provides an estimate of the open-loop amplitude ξ′ M (ω,n) and the open-loop phase estimate ξ′ P (ω,n).

[0214] The trained ML prediction model may be represented by the (optimized) parameters of an artificial neural network, such as a recurrent neural network (eg comprising one or more layers including gated recurrent units (GRUs), see, e.g., EP 4 033 784 A1).

[0215] Examples of multichannel hearing systems are Figure 3 The directional system is shown in Figure 1, where a single-channel signal is obtained and further passed through a gain function g (n) The M microphone signals are processed before processing.

[0216] Figure 3 An example of a multi-channel hearing system is shown. Figure 3 and Figure 1B Similar, but instead of an input transformer, Figure 3 The embodiment includes a plurality (M) of input transducers (eg, microphones M 1 ,…,M M ), each input converter experiences its own feedback path FBP from the output converter SPK to the input converter involved. Multiple feedback path transfer functions ( h 1 (n),…, h M (n)) may represent the impulse response of the corresponding feedback path FBP of the multi-channel hearing aid system. The multi-channel hearing system may be considered as a hearing aid comprising the multi-channel hearing system. For each of the M input transducers (feedback paths), a separate adaptive filter (ALG, FIL, h') for feedback estimation is implemented 1 (n),…,ALG,FIL, h ' M (n)).

[0217] Multi-channel hearing system operating in training mode

[0218] The multi-channel hearing system without the feedback cancellation system may be considered as a hearing aid operating in a training mode of operation. In other words, the simulation data may be provided by a computer simulation simulating the hearing aid system in a known, simulated acoustic environment (e.g., modeling a real-world condition or environment). While the hearing aid is operating in a normal operating mode, the ML prediction model used in the open-loop transfer function estimator OLTFE may be trained using the simulation data from the known, simulated acoustic environment. For example, in the training mode of operation, the weights of the ML prediction model may be updated based on the simulation data.

[0219] For example, simulation data includes the spatially filtered signal e(n), the processed signal, and multiple feedback path transfer functions (h 1 (n),…,h M(n)), an estimate of each of the multiple feedback path transfer functions (h' 1 (n),…,h' M (n)), and beamformer filters. For example, the spatially filtered signal (e(n)) refers to a signal applied to a plurality of electrical input signals (y 1 (n),…,y M (n)) of multiple signals (e 1 (n),…,e M (n)) of the beamformer filter BF. Dependent on multiple electrical input signals (y 1 (n),…,y M (n)) of multiple signals (e 1 (n),…,e M (n)) can be multiple feedback-corrected input signals, for example for a single-channel hearing aid system Figure 1B As set forth in . For example, each of the plurality of electrical input signals represents sound from a known, simulated acoustic environment of a hearing aid. For example, the processed output signal u(n) indicates that a frequency and / or level dependent gain function g(n) is applied to the spatially filtered signal (e(n)). The simulation data may include the frequency and / or level dependent gain function g(n). Optionally, the frequency and / or level dependent gain function g(n) may be inferred from the processed output signal u(n). For example, each of the plurality of feedback path transfer functions represents an impulse response of a corresponding feedback path FBP of the hearing aid.

[0220] Alternatively, for example for a multi-channel hearing system without a feedback cancellation system, the spatially filtered signal (e(n)) may indicate the application of the beamformer filter BF to the plurality of electrical input signals (y 1 (n),…,y M (n)).

[0221] For example, for a multi-channel hearing system including a feedback cancellation system, the spatially filtered signal (e(n)) may indicate the application of the beamformer filter BF to the multiple feedback input signals (e 1 (n),…,e M (n)).

[0222] Target open-loop transfer function ( Figure 7 of ) can be used to train the prediction model. For example, the target open-loop transfer function can be determined based on a gain function that varies with frequency and / or level, a plurality of feedback path transfer functions, and each of a plurality of estimators (an estimator of a corresponding feedback path transfer function in the plurality of feedback path transfer functions).

[0223] The target open-loop transfer function can be determined as:

[0224] ξ Targ (ω,n)=G(ω,n)·∑mB m (ω,n)·(H m (ω,n)-H' m (ω,n)) (3)

[0225] Where G(ω,n) is the frequency response of the gain function g(n) at time n, which varies with frequency and / or level, and H m (ω,n) refers to the frequency response of each of the multiple feedback path transfer functions at time n, H′ m (ω,n) refers to the frequency response of each of the multiple feedback path transfer functions at time n, B m (ω,n) refers to the frequency response of the beamformer filter for each input transducer channel in a plurality of input transducer channels (eg, each electrical input signal or each feedback-corrected input signal) at time n, ω refers to frequency, and (·) refers to the product operator.

[0226] For example, H m (ω,n) may refer to the frequency response of the mth feedback path transfer function, m = 1,…,M. For example, H′ m (ω,n) may refer to the frequency response of the estimate of the transfer function of the mth feedback path, m = 1,…,M. For example, B m (ω,n) refers to the mth frequency response of the beamformer filter, for example, for the mth input transducer channel (eg, the mth electrical input signal of the plurality of electrical input signals or the mth feedback-corrected input signal of the plurality of electrical input signals).

[0227] A training open-loop transfer function may be used to train the prediction model. For example, the training open-loop transfer function may be determined based on the spatially filtered signal e(n), the processed output signal u(n), and a gain function g(n) that varies with frequency and / or level.

[0228] For example, the ML prediction model is configured to receive simulation data as input and provide a trained open-loop transfer function as output.

[0229] For example, Figure 1A The system may include an open-loop transfer function estimator OLTFE in a training operation mode, which includes a configuration to provide a Figure 7 The training open-loop transfer function (For example, training open loop amplitude and training open loop phase )’s ML prediction model.

[0230] Figure 7 The target open-loop transfer function is Especially its magnitude and Phase It can be calculated according to equation (3). For example, the target open-loop transfer function may also be referred to as a true open-loop transfer function or a reference open-loop transfer function.

[0231] Multi-channel hearing system operating in normal operating mode

[0232] The multi-channel hearing system may be considered as a hearing system operating in a normal operating mode. In other words, an open-loop transfer function estimator OLTFE of a hearing aid comprising a multi-channel hearing system may comprise a trained ML prediction model (eg, the weights of the ML prediction model may be fixed) and is ready to be deployed in the hearing aid.

[0233] Figure 3 Embodiments of hearing aids including a multi-channel hearing system (e.g. a processor PRO or a feedback control system) may include, for example, a combination of the present invention. Figure 2A An open loop transfer function estimator OLTFE is described.

[0234] The open-loop transfer function estimator OLTFE includes a trained ML prediction model that estimates the open-loop transfer function ξ′(ω,n) based on test data from a real acoustic environment (e.g., a situation), such as the sound of a hearing aid environment. For example, the open-loop transfer function estimator OLTFE receives as input the spatially filtered signal e(n) and the processed signal u(n), and outputs an estimated open-loop transfer function ξ′(ω,n) (e.g., an estimated open-loop amplitude ξ′ M (ω,n) and the estimated open-loop phase ξ′ p (ω,n)) is provided as an output. An estimated open-loop transfer function ξ′(ω,n) (e.g., an estimated open-loop amplitude ξ′ M (ω,n) and the estimated open-loop phase ξ′ p (ω, n)) can be used to modify (e.g. adjust, update) the frequency- and / or level-dependent gain function g(n) of the signal processing unit PRO. If, for example, at certain frequencies ω, the open-loop amplitude approaches or exceeds 1 (0 dB) and / or the open-loop phase approaches 0 degrees, i.e., positive feedback may occur, the frequency- and / or level-dependent gain function g(n) can be changed so that its amplification at these frequencies ω will be reduced.

[0235] In other words, the frequency and / or level dependent gain function g(n) may be modified when the open loop amplitude approaches or exceeds 1 (0 dB) and / or the open loop phase approaches 0 degrees at certain frequencies ω. For example, the frequency and / or level dependent gain function g(n) may be modified based on an estimate that the open loop amplitude approaches or exceeds 1 (0 dB) and / or the open loop phase approaches 0 degrees at certain frequencies ω.

[0236] Optionally, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop amplitude at certain frequencies is approximately equal to or greater than an amplitude threshold (e.g., not necessarily 0 dB). For example, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop phase at certain frequencies approximately satisfies a threshold phase (e.g., not necessarily 0 degrees).

[0237] For example, the frequency and / or level dependent gain function g(n) may be modified when the estimated open loop amplitude is approximately equal to 0 dB. Alternatively, the frequency and / or level dependent gain function g(n) may be modified when the estimated open loop amplitude is within an amplitude range including a lower limit and an upper limit. The lower limit of the amplitude range may be approximately equal to -10 dB, -6 dB, -3 dB, or 0 dB. The upper limit of the amplitude range may be approximately equal to +3 dB, +6 dB, +10 dB, or a value greater than +10 dB. The amplitude range may include the following ranges: [-10dB, +3dB], [-10dB, +6dB], [-10dB, +10dB], [-10dB, > +10dB], [-6dB, +3dB], [-6dB, +6dB], [-6dB, +10dB], [-10dB, > +10dB], [-3dB, +3dB], [-3dB, +6dB], [-3dB, +10dB], [-3dB, > +10dB], [0dB, +3dB], [0dB, +6dB], [0dB, +10dB], [0dB, > +10dB]. For example, “> +10dB” may be considered a value greater than 10dB. For example, the amplitude range may include a range of [-20dB, 0dB].

[0238] For example, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop phase is approximately equal to 0 degrees. Alternatively, the frequency- and / or level-dependent gain function g(n) may be modified when the estimated open-loop phase is within a phase range. For example, the phase range may include the following ranges: [-180 degrees, +180 degrees], [-30 degrees, +30 degrees], [-60 degrees, 60 degrees], [-60 degrees, 30 degrees], and [-30 degrees, 60 degrees].

[0239] For example, the modification of the frequency- and / or level-dependent gain function g(n) may be different with frequency (e.g. because g (n) is a function of frequency).

[0240] The open loop transfer function estimator OLTFE may form part of a processor or a feedback control system (eg Figure 1B ) or can be separate units (such as Figure 1A as in ).

[0241] Figure 4A RITE hearing aid embodiment of a hearing aid according to the invention is shown. Figure 4 An embodiment of a hearing device HD according to the invention is shown. An exemplary hearing device HD, such as a hearing aid, is of a specific type (sometimes referred to as a receiver-in-the-ear or RITE type) comprising a BTE part (BTE) adapted to be located at or behind the ear of a user and an ITE part (ITE) adapted to be located in or at the ear canal of a user and comprising a receiver (speaker). The BTE part and the ITE part are connected (e.g. electrically connected) via a connection element IC and internal wiring in the ITE part and the BTE part (e.g. see wiring Wx in the BTE part). As an alternative, the connection element may be constituted entirely or partly by a wireless link between the BTE part and the ITE part (or, if the speaker is located in the BTE part, by a sound tube).

[0242] exist Figure 4 In an embodiment of the hearing device, the BTE part comprises a (first) input transducer (eg, microphone) having two (first) input transducers (eg, microphones) (M BTE1 ,M BTE2 ), each input transducer providing an input sound signal S representing the sound field S surrounding the hearing aid BTE The input unit also includes two wireless receivers (WLR 1 ,WLR 2 ) (or transceiver) for providing corresponding directly received auxiliary audio signals and / or control input signals (and / or enabling the transmission of audio and / or control signals to other devices such as another hearing device or to a remote control or processing device or a telephone). The hearing device HD comprises a substrate SUB on which a plurality of electronic components are mounted, including a memory MEM for storing, for example, different hearing aid programs (e.g. defining parameter settings of the aforementioned programs, or parameters of algorithms) and / or hearing aid configurations, such as input sound source combinations (MEM BTE1 ,M BTE2 ,M ITE,env ,M ITE,ed ,WLR 1 ,WLR 2 ), for example, optimized for a variety of different listening situations. In a particular operating mode, one or more directly received auxiliary electrical signals may be used together with one or more electrical input signals from the microphone to provide a beamforming signal, which is provided by applying appropriate complex weights to (at least parts of) the respective signals, for example, to provide an enhanced target signal to the user (or to provide an estimate of the user's own voice to another application such as a communication partner or a voice control interface).

[0243] The substrate SUB further comprises a configurable signal processor DSP (e.g. a digital (audio) signal processor), e.g. comprising a processor for applying a frequency and level dependent gain, e.g. providing hearing loss compensation, beam forming, noise reduction, filter bank functionality and other digital functions of the hearing device. The configurable signal processor DSP is adapted to access the memory MEM. The configurable signal processor DSP is further configured to process one or more electrical input audio signals and / or one or more directly received auxiliary audio input signals based on a currently selected (activated) (e.g. or automatically selected, e.g. selected based on one or more sensors, or based on input from a user interface) hearing aid program / parameter settings. The mentioned functional units (and other elements) may be divided in circuits and elements, 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. inductors, capacitors, etc.) depending on the application involved (e.g. for size, power consumption, analog-to-digital processing, acceptable latency, etc.). The configurable signal processor DSP provides a processed audio signal, which is intended to be presented to a user. The substrate also includes a front-end IC (FE) for interfacing a configurable signal processor DSP to input and output transducers, etc., typically including interfaces between analog and digital signals (e.g., interfaces to microphones and / or speakers). The input and output transducers can be individual separate components or integrated with other electronic circuits (e.g., based on MEMS).

[0244] The hearing device HD further comprises an output unit (eg an output transducer) for providing stimuli perceivable as sound by the user based on the processed audio signal from the processor or a signal derived therefrom. Figure 4 In an embodiment of a hearing device of the type described herein, the ITE part comprises an output transducer in the form of a loudspeaker (also referred to as a "receiver") SPK for converting an electrical signal into an acoustic (air-borne) signal which (when the hearing device is mounted at the ear of the user) is directed to the eardrum, where it provides a sound signal S ED The ITE part also includes a guide member such as a dome member DO for guiding and positioning the ITE part in the ear canal of the user. Figure 4 ) may also include an additional (first) input transducer such as a microphone (M ITE,env ), which faces the environment and is used to provide an input sound signal S representing the ear canal ITE The ITE part (such as Figure 4 ) may also include an additional (second) input transducer such as a microphone (M ITE,ed ), which is directed toward the eardrum and is used to provide a sound signal representing the eardrum (S ED =S dir +S HI) of the (second) electrical input audio signal. Sound S ITE The propagation from the environment via the direct acoustic path through the semi-open dome DO to the residual cavity at the eardrum is Figure 4 The direct path is indicated by the dotted arrow. dir mark) and the sound from the hearing device HD (via the sound field S HI The sound field S obtained at the eardrum is mixed with ED The sound output of the hearing device S HI Modifications may be made (at least in certain operating modes) based on the sound transmitted directly from the environment to the eardrum to provide adaptive noise cancellation (ANC) and / or adaptive occlusion control (AOC).

[0245] In addition to the (acoustic) output and input transducers, the ITE part may also include other functional elements such as (additional) detectors, for example electrodes for collecting signals from the user's body (such as brain wave signals, temperature indications, blood-related parameters, heartbeat indications, muscle vibrations). Such detectors may include one or more of the following: electroencephalogram (EEG) sensors, electromyogram (EMG) sensors, motion sensors, temperature sensors, photoplethysmogram (PPG) sensors, electrooculogram (EOG) sensors, etc.

[0246] (from (first and / or second) input converter M BTE1 ,M BTE2 ,M ITE,env ,M ITE,ed The electrical input signal may be processed in the time domain or in the (time-)frequency domain (or partly in the time domain and partly in the frequency domain, if deemed advantageous for the application in question).

[0247] Figure 4 The embodiment of a hearing device HD, such as a hearing aid, illustrated in the example is a portable device comprising a battery BAT, such as a rechargeable battery, e.g. based on lithium-ion battery technology, e.g. for powering the electronic components of the BTE part and possibly the ITE part. In an embodiment, the hearing device, such as a hearing aid, is adapted to provide a frequency-dependent gain and / or a level-dependent compression and / or a frequency shift of one or more frequency ranges to one or more other frequency ranges (with or without frequency compression), e.g. in order to compensate for a hearing impairment of the user. The BTE part may, for example, comprise a connector (e.g. a DAI or USB connector) for enabling a "boot" with additional functionality (e.g. an FM boot or an additional battery, etc.), a programming device, a charger, a separate processing device, etc. to be connected to the hearing device HD.

[0248] Figure 5A1 is a flow chart showing an exemplary method 100 for estimating an open-loop transfer function according to the present invention, performed by a hearing aid without a feedback cancellation system. The hearing aid without a feedback cancellation system is a hearing aid disclosed herein, for example Figure 1A hearing aids.

[0249] The method 100 comprises step S102 of obtaining an electric input signal by obtaining a sound representing an environment of the hearing aid. The hearing aid may obtain the electric input signal y(n) from an input unit of the hearing aid.

[0250] In one or more exemplary methods, at least one electrical input signal representing the sound of the hearing aid environment may be interpreted as a signal from a real sound environment, such as the sound environment of a user using the hearing aid (or where the hearing aid is in use). In other words, the hearing aid may be operated in a normal operating mode. The hearing aid may perform method 100 while operating in the normal operating mode.

[0251] The method 100 comprises step S104 of applying a frequency- and / or level-dependent gain function to the electrical input signal. The signal processing unit of the hearing aid may be configured to apply the aforementioned frequency- and / or level-dependent gain function to the electrical input signal.

[0252] The method 100 comprises a step S106 of providing a processed output signal in dependence on the applied frequency and / or level dependent gain function and the electrical input signal.The signal processing unit of the hearing aid may be configured to provide the processed output signal.

[0253] The method 100 includes a step S108 of estimating an open-loop transfer function based on the electrical input signal and the processed output signal. For example, the method 100 is a machine learning (ML) inference method. In other words, the estimated open-loop transfer function may be an inferred (e.g., derived) ML output. For example, estimating the open-loop transfer function may include applying the electrical input signal and the processed output signal to a trained ML model, such as a trained ML prediction model. An open-loop transfer function estimator (OLTFE) of the hearing aid (the open-loop transfer function estimator includes the trained ML prediction model) may be configured to estimate the open-loop transfer function. In other words, the trained ML prediction model may be configured to estimate the open-loop transfer function. In one or more exemplary methods, the step S108 of estimating the open-loop transfer function includes a step S108A of applying the electrical input signal and the processed output signal to the trained ML prediction model. Optionally, the step S108 of estimating the open-loop transfer function includes applying the electrical input signal and the processed output signal to an open-loop transfer function estimator (OLTFE) including the trained ML prediction model.

[0254] In one or more example methods, the estimated open-loop transfer function includes an estimated open-loop amplitude and an estimated open-loop phase. In one or more example methods, method 100 includes estimating the open-loop amplitude and the open-loop phase according to (eg, based on) the electrical input signal and the processed output signal.

[0255] In one or more exemplary methods, method 100 includes step S110 of controlling a frequency- and / or level-dependent gain function of the hearing aid according to the estimated open-loop transfer function (e.g., the estimated open-loop amplitude and the estimated open-loop phase). For example, method 100 includes adjusting (e.g., updating) the frequency- and / or level-dependent gain function of the hearing aid based on the estimated open-loop transfer function.

[0256] Figure 5B 1 is a flow chart showing an exemplary method 200 for estimating an open-loop transfer function according to the present invention, performed by a hearing aid comprising a feedback cancellation system. The hearing aid comprising a feedback cancellation system is a hearing aid disclosed herein, for example Figure 1B hearing aids.

[0257] The method 200 comprises step S202 of obtaining an electrical input signal by obtaining a sound representing an environment of the hearing aid. The hearing aid may obtain the electrical input signal from an input unit of the hearing aid.

[0258] In one or more exemplary methods, at least one electrical input signal representing the sound of the hearing aid environment may be interpreted as a signal from a real acoustic environment, such as the acoustic environment of a user using the hearing aid (or where the hearing aid is in use). In other words, the hearing aid may be operated in a normal operating mode. The hearing aid may perform method 200 while operating in the normal operating mode.

[0259] In one or more exemplary methods, the method 200 includes step S204 of determining a feedback-corrected input signal based on the electrical input signal and the estimated amount of the current feedback signal. The feedback-corrected input signal may refer to a signal in which acoustic or mechanical or electrical feedback is reduced or eliminated. The acoustic or mechanical or electrical feedback may originate from a feedback path from an output transducer of the hearing aid to an input unit. A feedback cancellation system of the hearing aid may be configured to determine the feedback-corrected input signal.

[0260] The current feedback signal may indicate a feedback path transfer function. For example, the current feedback signal includes a feedback path transfer function. The feedback path transfer function may represent an impulse response of the feedback path from the output converter to the input unit in the electrical input signal.

[0261] In one or more exemplary methods, the step S204 of determining the feedback-corrected input signal includes a step S204A of determining an estimate of the current feedback signal based on a previously determined feedback-corrected input signal and a previously determined processed signal.

[0262] The estimate of the current feedback signal may indicate an estimate of a transfer function of the feedback path. For example, the estimate of the current feedback signal includes an estimate of a transfer function of the feedback path. The estimate of the transfer function of the feedback path may represent an estimate of an impulse response of the feedback path.

[0263] For example, the feedback cancellation system may include an adaptive filter configured to provide an estimate of a current feedback signal, such as determining the estimate of the current feedback signal based on a previously determined feedback-corrected input signal and a previously determined processed signal. In other words, the adaptive filter may include a variable filter whose filter coefficients are determined (repeatedly updated) by an adaptive algorithm based on a previously determined feedback-corrected input signal and a previously determined processed signal. For example, the adaptive filter may be configured to provide an estimate of a feedback path transfer function.

[0264] The method 200 comprises step S206 of applying a frequency and / or level dependent gain function to the feedback corrected input signal. The signal processing unit of the hearing aid may be configured to apply the aforementioned frequency and / or level dependent gain function to the feedback corrected input signal.

[0265] The method 200 comprises step S208 of providing a processed output signal according to the applied frequency and / or level dependent gain function and the feedback corrected input signal.The signal processing unit of the hearing aid may be configured to provide the processed output signal.

[0266] The method 200 includes a step S210 of estimating an open-loop transfer function based on a feedback-corrected input signal and a processed output signal. For example, the method 200 is a machine learning (ML) inference method. In other words, the estimated open-loop transfer function may be an inferred (e.g., derived) ML output. For example, estimating the open-loop transfer function may include applying the electrical input signal and the processed output signal to a trained ML model, such as a trained ML prediction model. An open-loop transfer function estimator (OLTFE) of a hearing aid (the open-loop transfer function estimator including the trained ML prediction model) may be configured to estimate the open-loop transfer function. In other words, the trained ML prediction model may be configured to estimate the open-loop transfer function. In one or more exemplary methods, the step S210 of estimating the open-loop transfer function includes a step S210A of applying the feedback-corrected input signal and the processed output signal to the trained ML prediction model. Optionally, the step S210 of estimating the open-loop transfer function includes applying the feedback-corrected input signal and the processed output signal to an open-loop transfer function estimator (OLTFE) including the trained ML prediction model.

[0267] In one or more example methods, the estimated open-loop transfer function includes an estimated open-loop amplitude and an estimated open-loop phase. In one or more example methods, method 200 includes estimating the open-loop amplitude and the open-loop phase according to (eg, based on) the electrical input signal and the processed output signal.

[0268] In one or more exemplary methods, the method 200 includes step S212 of controlling a frequency- and / or level-dependent gain function of the hearing aid according to the estimated open-loop transfer function (e.g., the estimated open-loop amplitude and the estimated open-loop phase). For example, the method 200 includes adjusting (e.g., updating) the frequency- and / or level-dependent gain function of the hearing aid based on the estimated open-loop transfer function.

[0269] Figure 5C 1 is a flow chart showing an exemplary method 300 for estimating an open-loop transfer function according to the present invention, performed by a hearing aid comprising a multi-channel hearing aid system. The hearing aid comprising a multi-channel hearing aid system is a hearing aid disclosed herein, for example Figure 3 hearing aids.

[0270] The method 300 comprises step S302 of obtaining a plurality of electrical input signals representing the sound of the hearing aid environment. The hearing aid may obtain the plurality of electrical input signals from a corresponding plurality of input transducers (eg microphones) of the hearing aid.

[0271] In one or more exemplary methods, each of the plurality of electrical input signals representing sounds of the hearing aid environment may be interpreted as a signal from a real acoustic environment, such as an acoustic environment in which a user using the hearing aid is located (or in which the hearing aid is in use). In other words, the hearing aid may be operated in a normal operating mode. The hearing aid may perform method 300 while operating in a normal operating mode.

[0272] For example, the method 300 includes obtaining a first electrical input signal from a first input transducer of the plurality of input transducers. For example, the method 300 includes obtaining a second electrical input signal from a second input transducer of the plurality of input transducers. For example, the method 300 includes obtaining a third electrical input signal from a third input transducer of the plurality of input transducers.

[0273] In one or more exemplary methods, the method 300 includes step S304 of determining a plurality of feedback-corrected input signals, each of the plurality of feedback-corrected input signals being determined based on an estimate of a corresponding electrical input signal in a plurality of electrical input signals and a corresponding current feedback signal in a plurality of current feedback signals. Each of the plurality of feedback-corrected input signals may refer to a signal in which acoustic or mechanical or electrical feedback is reduced or eliminated. The acoustic or mechanical or electrical feedback may originate from a feedback path from an output transducer of the hearing aid to an input unit. The feedback cancellation system of the hearing aid may be configured to determine a plurality of feedback-corrected input signals.

[0274] Each of the plurality of current feedback signals may indicate a feedback path transfer function. For example, the current feedback signal includes a feedback path transfer function. The feedback path transfer function may represent an impulse response of a feedback path from the output converter to the input unit in the electrical input signal.

[0275] For example, the multiple current feedback signals include a first current feedback signal, a second current feedback signal, and a third current feedback signal.

[0276] In one or more example methods, method 300 includes determining an input signal for a first feedback correction. The input signal for the first feedback correction may be determined based on an estimate of the first electrical input signal and the first current feedback signal. In one or more example methods, method 300 includes determining an input signal for a second feedback correction. The input signal for the second feedback correction may be determined based on an estimate of the second electrical input signal and the second current feedback signal. In one or more example methods, method 300 includes determining an input signal for a third feedback correction. The input signal for the third feedback correction may be determined based on an estimate of the third electrical input signal and the third current feedback signal.

[0277] In one or more exemplary methods, the step S304 of determining a plurality of feedback-corrected input signals includes a step S304A of determining an estimate of each of the plurality of current feedback signals based on a corresponding previously determined feedback-corrected input signal and a corresponding previously determined processed signal.

[0278] The estimate of each of the plurality of current feedback signals may indicate an estimate of a feedback path transfer function. For example, the estimate of each of the plurality of current feedback signals includes an estimate of a feedback path transfer function. The estimate of the feedback path transfer function may represent an estimate of an impulse response of the feedback path.

[0279] For example, the feedback cancellation system may include a plurality of adaptive filters, each of the plurality of adaptive filters being configured to provide an estimate of a current feedback signal (e.g., determining an estimate of the current feedback signal based on a previously determined feedback-corrected input signal and a previously determined processed signal). In other words, each of the plurality of adaptive filters may include a variable filter whose adaptive algorithm and filter coefficients are determined (repeatedly updated) by the adaptive algorithm based on a previously determined feedback-corrected input signal and a previously determined processed signal. For example, each adaptive filter may be configured to provide an estimate of a feedback path transfer function.

[0280] In one or more exemplary methods, determining a first feedback-corrected input signal includes determining an estimate of a first current feedback signal based on a first previously determined feedback-corrected input signal and a first previously determined processed signal. A first adaptive filter of a plurality of adaptive filters may be configured to provide an estimate of the first current feedback signal. For example, the first adaptive filter may be configured to provide an estimate of a first feedback path transfer function, the first feedback path transfer function representing an estimate of an impulse response of a first feedback path from an output transducer to an input unit in a first electrical input signal.

[0281] In one or more exemplary methods, determining the second feedback-corrected input signal includes determining an estimate of the second current feedback signal based on the second previously determined feedback-corrected input signal and the second previously determined processed signal. For example, a second adaptive filter of the plurality of adaptive filters may be configured to provide an estimate of the second current feedback signal. For example, the second adaptive filter may be configured to provide an estimate of a second feedback path transfer function, the second feedback path transfer function representing an estimate of an impulse response of a second feedback path from the output transducer to the input unit in the second electrical input signal.

[0282] In one or more exemplary methods, determining the third feedback-corrected input signal includes determining an estimate of a third current feedback signal based on a third previously determined feedback-corrected input signal and a third previously determined processed signal. For example, a third adaptive filter of the plurality of adaptive filters may be configured to provide an estimate of the third current feedback signal. For example, the third adaptive filter may be configured to provide an estimate of a third feedback path transfer function, the third feedback path transfer function representing an estimate of an impulse response of a third feedback path from the output transducer to the input unit in the third electrical input signal.

[0283] The method 300 includes step S306 of determining a spatially filtered signal based on a plurality of feedback corrected input signals. In other words, the method 300 may include applying a beamformer filter to the plurality of feedback corrected input signals. The beamformer filter of the hearing aid may be configured to determine the spatially filtered signal. In one or more exemplary methods, the method includes determining the spatially filtered signal based on a first feedback corrected input signal, a second feedback corrected input signal, and a third feedback corrected input signal. The hearing aid may include a multi-hearing system, the multi-hearing system including a feedback cancellation system.

[0284] Optionally, the method 300 includes determining the spatially filtered signal based on the plurality of electrical input signals. In other words, the method 300 may include applying a beamformer filter to the plurality of electrical input signals. In one or more exemplary methods, the method 300 includes determining the spatially filtered signal based on the first electrical input signal, the second 3, and the third electrical input signal. The hearing aid may include a multi-hearing system without a feedback cancellation system.

[0285] The method 300 comprises step S308 of applying a frequency and / or level dependent gain function to the spatially filtered signal. The signal processing unit of the hearing aid may be configured to apply the aforementioned frequency and / or level dependent gain function to the spatially filtered signal.

[0286] The method 300 comprises a step S310 of providing a processed output signal according to the applied frequency and / or level dependent gain function and the spatially filtered signal.The signal processing unit of the hearing aid may be configured to provide the processed output signal.

[0287] The method 300 includes a step S312 of estimating an open-loop transfer function based on the spatially filtered signal and the processed output signal. For example, the method 300 is a machine learning (ML) inference method. In other words, the estimated open-loop transfer function may be an inferred (e.g., derived) ML output. For example, estimating the open-loop transfer function may include applying the spatially filtered signal and the processed output signal to a trained ML model, such as a trained ML prediction model. An open-loop transfer function estimator (OLTFE) of the hearing aid (the open-loop transfer function estimator includes the trained ML prediction model) may be configured to estimate the open-loop transfer function. In other words, the trained ML prediction model may be configured to estimate the open-loop transfer function. In one or more exemplary methods, the step S312 of estimating the open-loop transfer function includes a step S312A of applying the spatially filtered signal and the processed output signal to the trained ML prediction model. Optionally, the step S312 of estimating the open-loop transfer function includes applying the spatially filtered signal and the processed output signal to an open-loop transfer function estimator (OLTFE) including the trained ML prediction model.

[0288] In one or more exemplary methods, the estimated open-loop transfer function includes an estimated open-loop amplitude and an estimated open-loop phase. In one or more exemplary methods, method 300 includes estimating the open-loop amplitude and the open-loop phase based on (eg, based on) the spatially filtered signal and the processed output signal.

[0289] In one or more exemplary methods, method 300 includes step S314 of controlling a frequency- and / or level-dependent gain function of the hearing aid according to the estimated open-loop transfer function (e.g., the estimated open-loop amplitude and the estimated open-loop phase). For example, method 300 includes adjusting (e.g., updating) the frequency- and / or level-dependent gain function of the hearing aid based on the estimated open-loop transfer function.

[0290] Fig. 6A 4 is a flowchart showing an exemplary training method 400 of an ML prediction model for use in an open-loop transfer function estimator of a hearing aid according to the present invention. The hearing aid is a hearing aid without a feedback cancellation system (e.g. Figure 1A hearing aids).

[0291] The open loop transfer function estimator (OLFTE) includes an ML prediction model.

[0292] In one or more exemplary methods, the method 400 is performed by an external device (e.g., a computer). In other words, the method 400 may be a computer-implemented method for training a hearing device (e.g., Figure 1A Method for applying an ML model (e.g., a predictive model) to a hearing aid.

[0293] The training phase may be followed by an inference phase. In other words, the method 400 (e.g., training a prediction model) may be followed by an inference method such as Figure 5A Method 100. During the training phase, the weights associated with the ML prediction model may be (continuously) updated. In the inference phase, the ML prediction model is trained (eg, the weights may be fixed) and is ready for deployment.

[0294] The method 400 includes step S404 of performing a plurality of training iterations.

[0295] Each training iteration of the plurality of training iterations comprises step S404A of obtaining simulation data from a hearing aid.

[0296] The simulation data includes an electrical input signal, a processed output signal, and a feedback path transfer function. The electrical input signal represents sound from a known, simulated acoustic environment of the hearing aid. The processed output signal indicates a gain function that varies with frequency and / or level and is applied to the electrical input signal. The feedback path transfer function represents an impulse response of the feedback path of the hearing aid.

[0297] Each training iteration of the plurality of training iterations comprises step S404B of determining a target open-loop transfer function based on the gain function that varies with frequency and / or level and the feedback path transfer function.

[0298] Each training iteration of the plurality of training iterations comprises step S404C of determining a training open-loop transfer function based on the electrical input signal, the processed output signal, and the gain function that varies with frequency and / or level.

[0299] For example, the ML prediction model is configured to receive as input an electrical input signal, a processed output signal, and a gain function that varies with frequency and / or level, and to provide as output a trained open-loop transfer function.

[0300] Each training iteration of the multiple training iterations includes step S404D, updating the ML prediction model based on the target open-loop transfer function and the training open-loop transfer function.

[0301] In one or more exemplary methods, step S404B of determining a target open-loop transfer function includes step S404BA of determining a frequency response G(ω,n) of an applied frequency and / or level dependent gain function.

[0302] In one or more exemplary methods, the step S404B of determining the target open-loop transfer function includes a step S404BB of determining a frequency response H(ω,n) of the feedback path transfer function.

[0303] In one or more exemplary methods, step S404B of determining a target open-loop transfer function includes step S404BC of determining the target open-loop transfer function as G(ω,n)·H(ω,n), where n is time, ω is frequency, and (·) is a product operator.

[0304] In one or more exemplary methods, step S404C of determining a training open-loop transfer function includes step S404CA of providing the electrical input signal and the frequency and / or level-dependent gain function as inputs to the ML prediction model.

[0305] In one or more example methods, the ML prediction model includes one or more of the following: a deep neural network (DNN), a convolutional neural network (CNN), and a recurrent neural network (RNN).

[0306] In one or more exemplary methods, the ML prediction model includes a deep neural network (DNN). For example, the DNN may include at least two networks (e.g., layers). For example, the DNN may include one or more of the following: a convolutional neural network (CNN), and a recurrent neural network (RNN). For example, the DNN may include one or more of the following: a convolution-based neural network, a recurrent-based neural network. The RNN may include a gated recurrent unit (GRU).

[0307] In one or more exemplary methods, step S404D of updating the ML prediction model includes step S404DA of determining a training error signal based on the target open-loop transfer function and the training open-loop transfer function.

[0308] In one or more exemplary methods, step S404D of updating the ML prediction model includes step S404DB of updating the weights of the ML prediction model using a learning rule based on the training error signal.

[0309] In one or more exemplary methods, the training open-loop transfer function includes a training open-loop magnitude and a training open-loop phase. In one or more exemplary methods, the target open-loop transfer function includes a target open-loop magnitude and a target open-loop phase.

[0310] For example, the ML prediction model may be trained using an electrical input signal, a processed signal determined from the electrical input signal, a gain function that varies with frequency and / or level, and a feedback path transfer function.

[0311] Figure 6B 1 is a flow chart showing an exemplary training method 500 of an ML prediction model for use in an open-loop transfer function estimator of a hearing aid according to the present invention. The hearing aid is a hearing aid including a feedback cancellation system (e.g. Figure 1B hearing aids).

[0312] The open loop transfer function estimator (OLFTE) includes an ML prediction model.

[0313] In one or more exemplary methods, the method 500 is performed by an external device (e.g., a computer). In other words, the method 500 may be a computer-implemented method for training a hearing device (e.g., Figure 1B Method for applying an ML model (e.g., a predictive model) to a hearing aid.

[0314] The training phase may be followed by an inference phase. In other words, the method 500 (e.g., training a prediction model) may be followed by an inference method such as Figure 5B Method 200. During the training phase, the weights associated with the ML prediction model may be (continuously) updated. In the inference phase, the ML prediction model is trained (eg, the weights may be fixed) and is ready for deployment.

[0315] The method 500 includes step S504 of performing a plurality of training iterations.

[0316] Each training iteration of the plurality of training iterations comprises step S504A of obtaining simulation data from a hearing aid.

[0317] The simulation data includes an electrical input signal, a processed output signal, and a feedback path transfer function. The electrical input signal represents sound from a known, simulated acoustic environment of the hearing aid. The processed output signal indicates a gain function that varies with frequency and / or level applied to a signal derived from the electrical input signal. The feedback path transfer function represents an impulse response of the feedback path of the hearing aid.

[0318] In one or more exemplary methods, the simulation data also includes a feedback-corrected input signal and an estimate of a feedback path transfer function. In other words, a signal derived from the electrical input signal can be the feedback-corrected input signal. In one or more exemplary methods, the feedback-corrected input signal indicates a signal in which acoustic or mechanical or electrical feedback is reduced or eliminated, the acoustic or mechanical or electrical feedback originating from the feedback path. The electrical input signal can include the aforementioned acoustic or mechanical or electrical feedback originating from the feedback path, so that the feedback-corrected input signal is determined based on the electrical input signal.

[0319] In one or more exemplary methods, each training iteration of the plurality of training iterations includes step S504B of determining a target open-loop transfer function based on the frequency and / or level dependent gain function, the feedback path transfer function, and an estimate of the feedback path transfer function.

[0320] In one or more exemplary methods, each training iteration of the plurality of training iterations includes step S504C, determining a training open-loop transfer function based on a signal derived from an electrical input signal (e.g., a feedback-corrected input signal), a processed output signal, and a gain function that varies with frequency and / or level.

[0321] For example, the ML prediction model is configured to receive as input a signal derived from an electrical input signal (e.g., a feedback-corrected input signal), a processed output signal, and a gain function that varies with frequency and / or level, and provide as output a trained open-loop transfer function.

[0322] Each training iteration of the multiple training iterations includes step S504D, updating the ML prediction model based on the target open-loop transfer function and the training open-loop transfer function.

[0323] In one or more exemplary methods, step S504B of determining a target open-loop transfer function includes step S504BA of determining a frequency response G(ω,n) of the applied frequency and / or level dependent gain function.

[0324] In one or more exemplary methods, the step S504B of determining the target open-loop transfer function includes a step S504BB of determining a frequency response H(ω,n) of the feedback path transfer function.

[0325] In one or more exemplary methods, the step S504B of determining a target open-loop transfer function includes a step S504BC of determining a frequency response H′(ω,n) of an estimate of the feedback path transfer function.

[0326] In one or more exemplary methods, step S504B of determining a target open-loop transfer function includes step S504BD, determining the target open-loop transfer function as G(ω,n)·(H(ω,n)-H'(ω,n)), where n refers to time, ω refers to frequency, and (·) refers to a product operator.

[0327] In one or more exemplary methods, step S504C of determining a training open-loop transfer function includes step S504CA of providing a signal derived from an electrical input signal (e.g., a feedback-corrected input signal) and a gain function that varies with frequency and / or level as inputs to an ML prediction model.

[0328] In one or more exemplary methods, the ML prediction model includes a deep neural network (DNN). For example, the DNN may include at least two networks (e.g., layers). For example, the DNN may include one or more of the following: a convolutional neural network (CNN), and a recurrent neural network (RNN). For example, the DNN may include one or more of the following: a convolution-based neural network, a recurrent-based neural network. The RNN may include a gated recurrent unit (GRU).

[0329] In one or more exemplary methods, step S504D of updating the ML prediction model includes step S504DA of determining a training error signal based on the target open-loop transfer function and the training open-loop transfer function.

[0330] In one or more exemplary methods, step S504D of updating the ML prediction model includes step S504DB of updating the weights of the ML prediction model using a learning rule based on the training error signal.

[0331] In one or more exemplary methods, the training open-loop transfer function includes a training open-loop magnitude and a training open-loop phase. In one or more exemplary methods, the target open-loop transfer function includes a target open-loop magnitude and a target open-loop phase.

[0332] For example, the ML prediction model may be trained using a feedback-corrected input signal, a processed signal determined from 2, a gain function that varies with frequency and / or level, a feedback path transfer function, and an estimate of the feedback path transfer function.

[0333] Figure 6C 6 is a flowchart showing an exemplary training method 600 of an ML prediction model for use in an open-loop transfer function estimator of a hearing aid according to the present invention. The hearing aid is a hearing aid including a multi-channel system (e.g. Figure 3 hearing aids).

[0334] Hearing aids, including multi-channel systems, can be considered Figure 1A A hearing aid that includes multiple input transducers can be considered as a multi-channel system. Figure 1B The hearing aid includes a plurality of input transducers.

[0335] The open loop transfer function estimator (OLFTE) includes an ML prediction model.

[0336] In one or more exemplary methods, the method 600 is performed by an external device (e.g., a computer). In other words, the method 600 may be a computer-implemented method for training a hearing device (e.g., Figure 3 Method for applying an ML model (e.g., a predictive model) to a hearing aid.

[0337] The training phase may be followed by an inference phase. In other words, the method 600 (e.g., training a prediction model) may be followed by an inference method such as Figure 5C Method 300. During the training phase, the weights associated with the ML prediction model may be (continuously) updated. In the inference phase, the prediction model is trained (eg, the weights may be fixed) and ready for deployment.

[0338] The method 600 includes step S604 of performing a plurality of training iterations.

[0339] Each training iteration of the plurality of training iterations comprises step S604A of obtaining simulation data from a hearing aid.

[0340] The simulation data includes a plurality of electrical input signals, a processed output signal, and a plurality of feedback path transfer functions. Each electrical input signal represents sound from a known, simulated acoustic environment of a hearing aid. The processed output signal indicates a gain function that varies with frequency and / or level applied to a signal derived from the plurality of electrical input signals. The plurality of feedback path transfer functions represent impulse responses of a corresponding plurality of feedback paths of the hearing aid.

[0341] In one or more example methods, the plurality of electrical input signals may include M electrical input signals (eg, the hearing aid includes M input transducers).

[0342] For example, a signal derived from a plurality of electrical input signals may be considered as a spatially filtered signal. For example, the spatially filtered signal may indicate a beamformer filter applied to a plurality of electrical input signals (e.g. when the hearing aid does not include a feedback cancellation system). For example, the spatially filtered signal may indicate a beamformer filter applied to a plurality of electrical input signals, such as a plurality of feedback-corrected input signals (e.g. when the hearing aid includes a feedback cancellation system).

[0343] In one or more example methods, when the hearing aid does not include a feedback cancellation system, the simulation data may include a plurality of electrical input signals, a processed output signal, a plurality of feedback path transfer functions, and a spatially filtered signal.

[0344] In one or more example methods, when the hearing aid includes a feedback cancellation system, the simulation data may include multiple electrical input signals, processed output signals, multiple feedback path transfer functions, estimates of multiple feedback path transfer functions, and spatially filtered signals.

[0345] For example, the feedback corrected input signal indicates a signal in which acoustic or mechanical or electrical feedback originating from a feedback path is reduced or eliminated. Each electrical input signal may include acoustic or mechanical or electrical feedback originating from a corresponding feedback path of the plurality of feedback paths of the hearing aid.

[0346] In one or more exemplary methods, each of the plurality of training iterations includes step S604B of determining a target open-loop transfer function based on the frequency and / or level dependent gain function, the plurality of feedback path transfer functions, and estimates of the plurality of feedback path transfer functions.

[0347] In one or more exemplary methods, each of the multiple training iterations includes step S604C, determining a training open-loop transfer function based on a signal (e.g., a spatially filtered signal) derived from a plurality of electrical input signals, a processed output signal, and a gain function that varies with frequency and / or level.

[0348] For example, the ML prediction model is configured to receive as input a signal derived from a plurality of electrical input signals (e.g., a spatially filtered signal), a processed output signal, and a gain function that varies with frequency and / or level, and to provide as output a trained open-loop transfer function.

[0349] Each training iteration of the multiple training iterations includes step S604D, updating the ML prediction model based on the target open-loop transfer function and the training open-loop transfer function.

[0350] In one or more exemplary methods, step S604B of determining a target open-loop transfer function includes step S604BA of determining a frequency response G(ω,n) of the applied frequency and / or level dependent gain function.

[0351] In one or more exemplary methods, the step S604B of determining the target open-loop transfer function includes a step S604BB of determining a frequency response H of each of the plurality of feedback path transfer functions. m (ω,n). For example, determining the target open-loop transfer function includes determining the frequency response of the mth feedback path transfer function, m=1,…,M.

[0352] In one or more exemplary methods, the step S604B of determining the target open-loop transfer function includes a step S604BC of determining a frequency response H of an estimate of each of the plurality of feedback path transfer functions. ′ m (ω,n). For example, determining the target open-loop transfer function includes determining a frequency response of an estimate of the mth feedback path transfer function, m=1,…,M.

[0353] In one or more exemplary methods, the step S604B of determining the target open-loop transfer function includes a step S604BD of determining a frequency response B of a beamformer filter for each of the plurality of input transducer channels. m(ω, n). For example, determining the target open-loop transfer function includes determining the frequency response of the beamformer filter for the mth input transducer channel, m=1, ..., M. The term "input transducer channel" (e.g., microphone channel) in this specification means the input from a given input transducer (e.g., microphone) to the beamformer filter.

[0354] In one or more exemplary methods, the step S604B of determining the target open-loop transfer function includes a step S604BE of determining the target open-loop transfer function as G(ω,n)·∑mB m (ω,n)·(H m (ω,n)-H' m (ω,n)), where n is time, ω is frequency, and (·) is the product operator.

[0355] In one or more exemplary methods, step S604C of determining a training open-loop transfer function includes step S604CA of providing a signal derived from a plurality of electrical input signals (e.g., a spatially filtered signal), a gain function that varies with frequency and / or level, and a processed output signal as input to an ML prediction model.

[0356] In one or more exemplary methods, the ML prediction model includes a deep neural network (DNN). For example, the DNN may include at least two networks (e.g., layers). For example, the DNN may include one or more of the following: a convolutional neural network (CNN), and a recurrent neural network (RNN). For example, the DNN may include one or more of the following: a convolution-based neural network, a recurrent-based neural network. The RNN may include a gated recurrent unit (GRU).

[0357] In one or more exemplary methods, step S604D of updating the ML prediction model includes step S604DA of determining a training error signal based on the target open-loop transfer function and the training open-loop transfer function.

[0358] In one or more exemplary methods, step S604D of updating the ML prediction model includes step S604DB of updating the weights of the ML prediction model using a learning rule based on the training error signal.

[0359] In one or more exemplary methods, the training open-loop transfer function includes a training open-loop magnitude and a training open-loop phase. In one or more exemplary methods, the target open-loop transfer function includes a target open-loop magnitude and a target open-loop phase.

[0360] For example, the ML prediction model can use multiple electrical input signals (y 1 (n),…,y M (n)) or rely on multiple electrical input signals (y 1 (n),…,y M(n)) of multiple signals (e 1 (n),…,e M (n)), a processed signal determined from a plurality of electrical input signals or a plurality of signals dependent on a plurality of electrical input signals, a gain function that varies with frequency and / or level, a plurality of feedback path transfer functions ((h 1 (n),…,h M (n)), an estimate of each of the multiple feedback path transfer functions ((h' 1 (n),…,h' M (n)), and beamformer filters are trained.

[0361] Figure 7 An exemplary structure of an ML prediction model according to the present invention is schematically shown.

[0362] The open-loop transfer function estimator of the hearing aid comprises a ML prediction model (ML-PM). The ML prediction model is configured to receive simulation data as input and provide a trained open-loop transfer function as output.

[0363] For hearing aids without feedback cancellation systems

[0364] The simulation data may include an electrical input signal (y(n)), a processed output signal (u(n)), a gain function (g(n)) that varies with frequency and / or level, and a feedback path transfer function ( h (n)).

[0365] The electrical input signal (y(n)) represents the sound from a known, simulated acoustic environment of the hearing aid. The processed output signal (u(n)) specifies the frequency and / or level dependent gain function (g(n)) applied to the electrical input signal (y(n)). The feedback path transfer function ( h (n)) represents the impulse response of the feedback path (FBP) of the hearing aid.

[0366] The ML prediction model (e.g., ML model) (ML-PM) may be configured to determine a training open-loop transfer function based on an electrical input signal (y(n)), a processed output signal (u(n)), and a gain function (g(n)) that varies with frequency and / or level

[0367] Loss function (LF) can be configured to receive the target open loop transfer function Target open-loop transfer function Based on the frequency and / or level dependent gain function (g(n)) and the feedback path transfer function ( h (n)) determine (e.g., combine Fig. 6A described).

[0368] For hearing aids that include feedback cancellation systems

[0369] The simulation data may include the feedback-corrected input signal (e(n)), the processed output signal (u(n)), the gain function (g(n)) that varies with frequency and / or level, the feedback path transfer function ( h (n)), and the feedback path transfer function ( h (n)) h '(n)).

[0370] The feedback corrected input signal indicates the signal for which acoustic or mechanical or electrical feedback is reduced or eliminated, the acoustic or mechanical or electrical feedback originating from the feedback path (FBP). The processed output signal (u(n)) indicates the frequency and / or level dependent gain function (g(n)) applied to the feedback corrected input signal (e(n)). The feedback path transfer function ( h (n)) represents the impulse response of the feedback path (FBP) of the hearing aid.

[0371] The ML prediction model (ML-PM) can be configured to determine a training open-loop transfer function based on a feedback-corrected input signal (e(n)), a processed output signal (u(n)), and a gain function (g(n)) that varies with frequency and / or level

[0372] Loss function (LF) can be configured to receive the target open loop transfer function Target open-loop transfer function Based on the frequency and / or level dependent gain function (g(n)), feedback path transfer function ( h (n)), and the feedback path transfer function ( h (n)) h '(n)) determine (e.g. combine Figure 6B described).

[0373] For hearing aids that include a multi-channel system, the multi-channel system does not include a feedback cancellation system

[0374] The simulation data may include the spatially filtered signal (e(n)), the processed output signal (u(n)), the gain function (g(n)) that varies with frequency and / or level, and multiple feedback path transfer functions (h 1 (n),…,h M (n)).

[0375] The spatially filtered signal (e(n)) may indicate the signal applied to the plurality of electrical input signals (y 1 (n),…,y M (n)) of the beamformer filter, multiple electrical input signals (y 1 (n),…,y MEach of the feedback paths (n) represents a sound from a known, simulated acoustic environment of the hearing aid. The processed output signal (u(n)) indicates a frequency and / or level dependent gain function (g(n)) applied to the feedback spatially filtered signal (e(n)). The multiple feedback path transfer functions (h 1 (n),…,h M Each of (n)) represents an impulse response of a corresponding plurality of feedback paths of the hearing aid.

[0376] The ML prediction model (ML-PM) can be configured to determine a training open-loop transfer function based on the spatially filtered signal (e(n)), the processed output signal (u(n)), and the gain function (g(n)) that varies with frequency and / or level

[0377] Loss function (LF) can be configured to receive the target open loop transfer function Target open-loop transfer function Based on a frequency and / or level dependent gain function (g(n)) and multiple feedback path transfer functions (h 1 (n),…,h M (n)) determine (e.g., combine Figure 6C described).

[0378] For hearing aids that include a multi-channel system, the multi-channel system includes a feedback cancellation system

[0379] The simulation data may include the spatially filtered signal (e(n)), the processed output signal (u(n)), the gain function that varies with frequency and / or level (g(n)), the multiple feedback path transfer functions (h(n)), and the feedback path transfer functions (h(n)). 1 (n),…,h M (n)), and multiple feedback path transfer functions (h 1 (n),…,h M The estimate (h') of each of the 1 (n),…,h' M (n)).

[0380] The spatially filtered signal (e(n)) may indicate the input signal (e 1 (n),…,e M (n)) of the beamformer filter, the plurality of feedback-corrected input signals being based on the corresponding plurality of electrical input signals (y 1 (n),…,y M (n)) to determine. Multiple electrical input signals (y 1 (n),…,y MEach of the feedback paths (n) represents a sound from a known, simulated acoustic environment of the hearing aid. The processed output signal (u(n)) indicates a frequency and / or level dependent gain function (g(n)) applied to the feedback spatially filtered signal (e(n)). The multiple feedback path transfer functions (h 1 (n),…,h M Each of (n)) represents an impulse response of a corresponding plurality of feedback paths of the hearing aid.

[0381] The ML prediction model (ML-PM) can be configured to determine a training open-loop transfer function based on the spatially filtered signal (e(n)), the processed output signal (u(n)), and the gain function (g(n)) that varies with frequency and / or level

[0382] Loss function (LF) can be configured to receive the target open loop transfer function Target open-loop transfer function Based on frequency and / or level dependent gain functions (g(n)), multiple feedback path transfer functions (h 1 (n),…,h M (n)), multiple feedback path transfer functions (h 1 (n),…,h M The estimate (h') of each of the 1 (n),…,h' M (n)), and the beamformer filter determines (e.g., combining Figure 6C described).

[0383] For hearing aids without feedback cancellation systems, hearing aids with feedback cancellation systems, including multi-channel systems Hearing aids including multi-channel systems (multi-channel systems without feedback cancellation systems) Feedback Elimination System)

[0384] The ML prediction model (ML-PN) includes a deep neural network (DNN). For example, the DNN may include at least two neural networks (e.g., layers). For example, the DNN may include one or more of the following: a convolutional neural network (CNN), and a recurrent neural network (RNN). For example, the DNN may include one or more of the following: a convolution-based neural network, a recurrent-based neural network. The RNN may include a gated recurrent unit (GRU).

[0385] The loss function (LF) can be configured to be based on the target open loop transfer function and training open-loop transfer function Determine the training error signal (e T ).

[0386] The ML prediction model (eg, ML model) (LA-PM) may be configured to update weights using a learning rule based on a training error signal.

[0387] Training the open-loop transfer function Can include training open loop amplitude and training open loop phase Target open-loop transfer function Target open loop amplitude may be included and the target open-loop phase

[0388] The term "or a processed version thereof" may for example cover features extracted from an original audio signal. The term "or a processed version thereof" may for example also cover an original audio signal that has been subjected to a processing algorithm that applies a gain or attenuation and / or a delay to the original audio signal, which results in a modified audio signal (preferably enhanced in some aspect, e.g. noise reduced relative to a target signal, e.g. feedback correction, or simply delayed).

[0389] The structural features of the apparatus described above, described in detail in the “Detailed Description of the Invention” and defined in the claims may be combined with the steps of the method of the present invention when appropriately replaced by corresponding processes.

[0390] Unless expressly stated, the singular forms "one", "the" used herein include the plural form (i.e., have the meaning of "at least one"). It should be further understood that the terms "having", "including" and / or "comprising" used in the specification indicate the presence of the described features, integers, steps, operations, elements and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, parts and / or combinations thereof. It should be understood that, unless expressly stated, when an element is referred to as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be an intermediate intervening element. As used herein, the term "and / or" includes any and all combinations of one or more listed related items. Unless expressly stated, the steps of any method disclosed herein do not have to be performed in the exact order disclosed.

[0391] It should be appreciated that reference to "an embodiment" or "embodiment" or "aspect" or features that "may" include in this specification means that the specific features, structures or characteristics described in conjunction with the embodiment are included in at least one embodiment of the present invention. In addition, the specific features, structures or characteristics may be appropriately combined in one or more embodiments of the present invention. The foregoing description is provided to enable those skilled in the art to implement the various aspects described herein. Various modifications will be apparent to those skilled in the art.

[0392] The claims are not limited to the various aspects shown herein, but rather have the full scope consistent with the claim language wherein, unless expressly stated otherwise, elements referred to in the singular do not mean "one and only one" but rather "one or more." Unless expressly stated otherwise, the term "some" means one or more.

[0393] Examples of methods and products (methods and hearing aids) according to the invention are presented in the following projects:

[0394] Item 1. A hearing aid (HD) comprising a forward path for processing electrical signals representing sound, the forward path comprising:

[0395] - an input unit (IU) for receiving or providing at least one electrical input signal (y(n)) representing a sound;

[0396] - a signal processing unit (PRO) configured to convert a frequency and / or level-dependent gain function ( g (n)) applied to at least one electrical input signal (y(n)) or a signal derived therefrom and providing a processed output signal (u(n)) therefrom; and

[0397] - an output transducer (OT) for generating a stimulus perceivable as sound to a user based on the processed output signal (u(n));

[0398] The hearing aid (HD) further comprises an open-loop transfer function estimator (OLTFE) for estimating a current open-loop transfer function, wherein the current open-loop transfer function comprises an open-loop amplitude and an open-loop phase; and

[0399] The open-loop transfer function estimator includes a learning algorithm including a trained prediction model configured to provide an estimate of the open-loop amplitude (ξ' M (ω,n)) and the open-loop phase estimate (ξ' P (ω,n));

[0400] Therein, the hearing aid is configured in a training operating mode to enable training of the prediction model using simulated data from known, simulated acoustic situations.

[0401] Item 2. A hearing aid according to Item 1, wherein the prediction model has been trained using the simulation data, wherein the simulation data includes a processed signal (u(n)) determined from known input data (y(n)), a frequency and / or level dependent gain function ( g (n)), based on a gain function that varies with frequency and / or level ( g (n)) and the known feedback path transfer function ( h (n)) Calculate the open-loop amplitude and open-loop phase.

[0402] Item 3. A hearing aid according to Item 1 or 2, wherein the prediction model has been trained using simulation data from a plurality of simulations, the plurality of simulations comprising different feedback path transfer functions ( h (n)), different frequency and / or level dependent gain functions ( g (n)), and a different external input signal (x(n)) to the hearing aid, where the external input signal is the part of the electrical input signal (y(n)) that is not caused by feedback.

[0403] Item 4. The hearing aid of any one of items 1-3, wherein for the simulation data, the calculated open-loop transfer function is determined as

[0404] ξ(ω,n)=G(ω,n)·H(ω,n)

[0405] The calculated open-loop transfer function includes the calculated open-loop magnitude and phase, where G(ω,n) is a gain function that varies with frequency and / or level ( g (n)), H(ω,n) is the known feedback path transfer function ( h (n)), where n is time and is frequency.

[0406] Item 5. A hearing aid according to any one of items 1 to 4, further comprising a feedback control system configured to transfer the function ( h (n)) eliminates or reduces feedback in at least one electrical input signal (y(n)), provides an estimate (v'(n)) of a current feedback signal (v(n)) received by the input unit via the feedback path, and provides a feedback-corrected input signal (e(n)) based on at least one electrical input signal (y(n)) or a signal derived therefrom and the estimate (v'(n)) of the current feedback signal; wherein the feedback control system is configured to provide a feedback path impulse response ( h (n)) h '(n)).

[0407] Item 6. The hearing aid of Item 5, wherein for the simulation data, the calculated open-loop transfer function is determined as

[0408] ξ(ω,n)=G(ω,n)·(H(ω,n)-H'(ω,n))

[0409] where G(ω,n) is a gain function that varies with frequency and / or level ( g (n)), H(ω,n) is the known feedback path transfer function ( h(n)) is the frequency response, and H'(ω,n) is the impulse response of the feedback path ( h (n)) h '(n)), where n is time and is frequency.

[0410] Item 7. A hearing aid according to Item 5 or 6, wherein the prediction model has been trained using the simulation data, wherein the simulation data includes a feedback-corrected input signal (e(n)) determined from known input data (y(n)), a processed signal (u(n)) determined from the known input data (y(n)), a frequency and / or level-dependent gain function ( g (n)), based on a gain function that varies with frequency and / or level ( g (n)), the known feedback path transfer function ( h (n)) and the known feedback path transfer function ( h (n)) h '(n)) to calculate the open-loop amplitude and open-loop phase.

[0411] Item 8. A hearing aid according to any one of Items 5-7, wherein the feedback control system comprises a control system configured to provide a current feedback signal ( h (n)) h '(n))'s adaptive filter (ALG,FIL, h '(n)).

[0412] Item 9. A hearing aid according to any one of items 1 to 8, comprising a beamformer filter (BF) configured to generate a beamformer filter (BF) according to a plurality of electrical input signals (y 1 (n),…,y M (n)) or a signal derived therefrom (e 1 (n),…,e M (n)) provides a spatially filtered signal (e(n)).

[0413] Item 10. The hearing aid of Item 9, wherein the open loop transfer function is given by:

[0414] ξ(ω,n)=G(ω,n)·∑ m B m (n)·(H m (ω,n)-H' m (ω,n)

[0415] Among them, B m (ω,n) is the frequency response of the beamformer filter for each microphone channel, m = 1,…,M,H m (ω,n) and H'm (ω,n) are the mth feedback paths ( h m (n)) and adaptive filter ( h ' m (n)) frequency response.

[0416] Item 11. A hearing aid according to any one of items 1-10, wherein the gain function ( g (n)) Control is performed based on an estimate of the open-loop transfer function.

[0417] Item 12. The hearing aid according to any one of items 1 to 11, which is composed of an air conduction hearing aid, a bone conduction hearing aid, or a combination thereof, or includes an air conduction hearing aid, a bone conduction hearing aid, or a combination thereof.

[0418] Item 13. A training method for an open-loop transfer function estimator of a hearing aid, the hearing aid (HD) comprising a forward path for processing an electrical signal representing sound, the forward path comprising:

[0419] - an input unit (IU) for receiving or providing at least one electrical input signal (y(n)) representing a sound;

[0420] - a signal processing unit (PRO) configured to convert a frequency and / or level-dependent gain function ( g (n)) applied to at least one electrical input signal (y(n)) or a signal derived therefrom and providing a processed output signal (u(n)) therefrom; and

[0421] - an output transducer (OT) for generating a stimulus perceivable as sound to a user based on the processed output signal (u(n));

[0422] wherein the hearing aid further comprises an open loop transfer function estimator (OLTFE) for estimating a current open loop transfer function in the form of open loop amplitude and open loop phase; and

[0423] Wherein, the method comprises:

[0424] - causing the open-loop transfer function estimator to include a learning algorithm including a prediction model configured to provide an open-loop amplitude and an open-loop phase;

[0425] Therein, the prediction model is trained using simulation data from known, simulated acoustic situations.

[0426] Item 14. A method according to Item 13, wherein the prediction model is trained using the simulation data, wherein the simulation data includes a feedback-corrected input signal (e(n)) determined from known input data (y(n)), a processed signal (u(n)) determined from the known input data (y(n)), a gain function ( g (n)), based on a gain function that varies with frequency and / or level ( g (n)) and the feedback path transfer function ( h (n)) and the feedback path transfer function ( h (n)) h '(n)) to calculate the open-loop amplitude and open-loop phase of one or more of the above.

[0427] Item 15. A method according to Item 13 or 14, wherein the prediction model is trained using simulation data from multiple simulations, the multiple simulations including different feedback path transfer functions ( h (n)), different frequency and / or level dependent gain functions ( g (n)), and different external input signals (x(n)) to the hearing aid, where the external input signal is the part of the electrical input signal that is not caused by feedback.

[0428] References

[0429] ·EP3291581A2(Oticon)07.03.2018.

[0430] ·EP4033784A1(Oticon)27.02.2022.

Claims

1. A method for training a machine learning (ML) prediction model in an open loop transfer function estimator (OLFTE) of a hearing aid (HD), wherein the open loop transfer function estimator (OLFTE) comprises a ML prediction model (ML-PM), wherein the method comprises performing a plurality of training iterations, each training iteration of the plurality of training iterations comprising: - obtaining simulation data from the hearing aid, the simulation data comprising - at least one electrical input signal (y(n)) representing sound from a known, simulated acoustic environment of a hearing aid (HD); - a processed output signal (u(n)) indicative of a frequency and / or level dependent gain function (g(n)) applied to at least one electrical input signal (y(n)) or a signal derived therefrom; and --Feedback path transfer function representing the impulse response of the hearing aid's feedback path (FBP) h (n)); and - Based on the frequency and / or level dependent gain function (g(n)) and the feedback path transfer function ( h (n)) Determine the target open-loop transfer function - determining a training open-loop transfer function based on at least one electrical input signal (y(n)) or a signal derived therefrom, a processed output signal (u(n)), and a gain function (g(n)) that varies with frequency and / or level and -According to the target open loop transfer function and train the open-loop transfer function The ML prediction model is updated.

2. The method according to claim 1, wherein: Determine the target open-loop transfer function include: - determining the frequency response G(ω,n) of the applied frequency and / or level dependent gain function (g(n)); - Determine the feedback path transfer function ( h (n))’s frequency response H(ω,n); and -The target open-loop transfer function Determined as Here, n refers to time and ω refers to frequency.

3. The method according to claim 1, wherein: The simulation data also includes: - a feedback corrected input signal (e(n)) indicative of a signal for reducing or eliminating acoustic or mechanical or electrical feedback originating from the feedback path (FBP); and -Feedback path transfer function ( h (n)) h '(n)); Each training iteration in the multiple training iterations includes: - Based on the frequency and / or level dependent gain function (g(n)), feedback path transfer function ( h (n)), and the feedback path transfer function ( h (n)) h '(n)) Determine the target open-loop transfer function and - Determine the training open-loop transfer function based on the feedback-corrected input signal (e(n)), the processed output signal (u(n)), and the gain function (g(n)) that varies with frequency and / or level 4. The method according to claim 3, wherein: Determine the target open-loop transfer function include: - determining the frequency response G(ω,n) of the applied frequency and / or level dependent gain function (g(n)); - Determine the feedback path transfer function ( h (n))’s frequency response H(ω,n); - Determine the feedback path transfer function ( h (n)) h '(n))’s frequency response H′(ω,n); and -The target open-loop transfer function Determined as Here, n refers to time and ω refers to frequency.

5. The method according to claim 1, wherein: Determine the training open-loop transfer function include: - providing as input to the ML prediction model at least one electrical input signal (y(n)) or a signal derived therefrom, a processed output signal (u(n)), and a frequency and / or level dependent gain function (g(n)).

6. The method according to claim 1, wherein: The ML prediction model includes a deep neural network.

7. The method according to claim 1, wherein: Updating the ML prediction model includes: -According to the target open loop transfer function and train the open-loop transfer function determining a training error signal; and - Updating the weights of the ML prediction model using a learning rule based on a training error signal.

8. The method according to claim 1, wherein: Training the open-loop transfer function Including training open loop amplitude and training open loop phase 9. The method according to claim 1, wherein: The method is executed by an external device.

10. A hearing aid (HD) comprising a forward path for processing electrical signals representing sound, the forward path comprising: - an input unit (IU) for receiving or providing at least one electrical input signal (y(n)) representing sounds of the hearing aid environment; - a signal processing unit (PRO) configured to convert a frequency and / or level-dependent gain function ( g (n)) applied to at least one electrical input signal (y(n)) or a signal derived therefrom and providing a processed output signal (u(n)) therefrom; and - an output transducer (OT) for generating a stimulus perceivable as sound to a user based on the processed output signal (u(n)); wherein the hearing aid (HD) further comprises an open-loop transfer function estimator (OLTFE) comprising a trained ML prediction model, configured to estimate an open-loop transfer function (ξ′(ω,n)) based on at least one electrical input signal (y(n)) or a signal derived therefrom and a processed output signal (u(n)); and Wherein, the ML prediction model is trained according to the method of claim 1.

11. The hearing aid according to claim 10, wherein: The hearing aid further comprises a feedback control system configured to eliminate or reduce feedback in at least one electrical input signal (y(n)) via an acoustic or mechanical or electrical feedback path (FBP) from the output transducer (OT) to the input unit (IU) and to provide: - an estimate (v'(n)) of the current feedback signal (v(n)) received via the input unit via said feedback path; - an input signal (e(n)) corrected by feedback of at least one electrical input signal (y(n)) or a signal derived therefrom and an estimate (v'(n)) of the current feedback signal (v(n)); and -Feedback path transfer function ( h (n)) h '(n)), where the feedback path transfer function ( h (n)) represents an impulse response of a feedback path (FBP) from the output transducer (OT) to the input unit (IU) in at least one electrical input signal (y(n)).

12. The hearing aid according to claim 11, wherein: The feedback control system includes a feedback path transfer function ( h (n)) h '(n))'s adaptive filter (ALG,FIL, h '(n)).

13. The hearing aid according to claim 11, wherein: An open-loop transfer function estimator (OLTFE) including a trained ML prediction model is configured to estimate an open-loop transfer function (ξ′(ω,n)) based on a feedback-corrected input signal (e(n)) and a processed output signal (u(n)).

14. The hearing aid according to claim 10, wherein: The estimated open-loop transfer function (ξ′(ω,n)) includes the estimated open-loop amplitude (ξ′ M (ω,n)) and the estimated open-loop phase (ξ′ p (ω,n)).

15. The hearing aid according to claim 10, wherein: The frequency and / or level dependent gain function (g(n)) is controlled based on the estimated open loop transfer function (ξ′(ω,n)).

Images