Search

CN-122002201-A - Hearing aid signal processing method, system, medium and product based on electroencephalogram detection

CN122002201ACN 122002201 ACN122002201 ACN 122002201ACN-122002201-A

Abstract

The invention discloses a hearing aid signal processing method, a system, a medium and a product based on electroencephalogram detection, belonging to the field of hearing aid signal processing, wherein the method comprises the steps of synchronously collecting original electroencephalogram signals, microphone array signals, inertial measurement data and historical previous frame audio driving signals of a user; performing blind source separation on the microphone array signals to obtain independent candidate sound source streams, performing motion artifact cancellation and audio interference cancellation processing on the original electroencephalogram signals according to inertia measurement data and the historical previous frame of audio driving signals to obtain electroencephalogram signals, performing neural decoding matching analysis on the independent candidate sound source streams and the electroencephalogram signals to obtain target sound source probability values, performing space enhancement beam forming processing according to the target sound source probability values and the microphone array signals to obtain enhanced audio signals, and outputting the enhanced audio signals. By implementing the invention, the problems of brain electrical signal distortion and hearing space loss caused by physical interference and algorithm limitation when the existing hearing aid is used for introducing brain electrical control can be solved.

Inventors

  • WANG PEI
  • GUO ZHIHAO
  • MIAO JIANZHANG

Assignees

  • 爱听智能科技(深圳)有限公司

Dates

Publication Date
20260508
Application Date
20260408

Claims (10)

  1. 1. A hearing aid signal processing method based on electroencephalogram detection, comprising: synchronously acquiring an original brain electrical signal, a microphone array signal, inertial measurement data and a historical previous frame of audio driving signal of a user through a hearing aid; performing blind source separation on the microphone array signals to obtain independent candidate sound source streams, and performing motion artifact cancellation and audio interference cancellation on the original brain electrical signals according to the inertia measurement data and the historical previous frame of audio driving signals to obtain brain electrical signals; and performing neural decoding matching analysis according to the independent candidate sound source flow and the electroencephalogram signal to obtain a target sound source probability value, performing space enhancement beam forming processing according to the target sound source probability value and the microphone array signal to obtain an enhancement audio signal and outputting the enhancement audio signal.
  2. 2. The method for processing a hearing aid signal based on electroencephalogram detection according to claim 1, wherein performing blind source separation on the microphone array signal to obtain independent candidate sound source streams comprises: According to the microphone array signals, calculating a received signal correlation matrix and a noise correlation matrix to obtain a statistical domain feature matrix; Identifying acoustic guide vectors of each independent sound source through a generalized eigenvalue decomposition algorithm according to the statistical domain eigenvalue matrix to obtain sound source space eigenvectors; Constructing a mutual exclusion space filter weight vector meeting the constraint conditions of unit gain and null linearity according to the sound source space feature vector to obtain a space filtering parameter; And based on the spatial filtering parameters, performing linear filtering operation on the microphone array signals to obtain independent candidate sound source streams corresponding to potential sound sources with different directions.
  3. 3. The method for processing hearing aid signals based on electroencephalogram detection according to claim 1, wherein said performing motion artifact cancellation and audio interference cancellation on the original electroencephalogram signal according to the inertial measurement data and the historical previous frame of audio driving signal to obtain an electroencephalogram signal comprises: constructing a first adaptive filter by using a recursive least square algorithm according to the inertial measurement data, estimating a motion artifact noise estimation signal related to the inertial measurement data, and filtering the motion artifact noise estimation signal from the original electroencephalogram signal to obtain an intermediate electroencephalogram signal; And constructing a second adaptive filter according to the historical previous frame of audio driving signal by utilizing a normalized least mean square algorithm, estimating an audio interference noise estimation signal related to the historical previous frame of audio driving signal, and filtering the audio interference noise estimation signal in the intermediate state electroencephalogram signal to obtain the electroencephalogram signal.
  4. 4. A hearing aid signal processing method based on electroencephalogram detection according to claim 3, wherein said constructing a first adaptive filter from said inertial measurement data using a recursive least squares algorithm, estimating a motion artifact noise estimation signal associated with said inertial measurement data and filtering out said raw electroencephalogram signal to obtain an intermediate electroencephalogram signal, comprises: initializing discrete time steps based on the duration of the original electroencephalogram signals, and initializing an inverse correlation matrix and a filter weight vector of a recursive least square algorithm; Performing recursive least square iteration processing according to the inertial measurement data and the original electroencephalogram until all discrete time steps are traversed, and outputting a final intermediate electroencephalogram; The method comprises the steps of extracting triaxial acceleration and triaxial angular velocity data before a current time step in inertia measurement data and constructing a first reference vector during each round of iterative processing, calculating a gain vector by using a current inverse correlation matrix, a preset forgetting factor and the first reference vector, combining a current filter weight vector and the first reference vector, estimating to obtain a motion artifact noise estimation signal of the current time step, subtracting the motion artifact noise estimation signal from an original electroencephalogram signal of the current time step to obtain an intermediate electroencephalogram signal of the current time step and serve as an priori error of a current round, and updating the current filter weight vector and the inverse correlation matrix according to the gain vector, a conjugate signal of the priori error and the forgetting factor.
  5. 5. A hearing aid signal processing method based on electroencephalogram detection as claimed in claim 3, wherein said constructing a second adaptive filter from said historical previous frame audio driving signal using a normalized least mean square algorithm, estimating an audio interference noise estimation signal associated with said historical previous frame audio driving signal, and filtering in said intermediate electroencephalogram signal to obtain said electroencephalogram signal, comprises: initializing discrete time steps based on the duration of the intermediate state electroencephalogram signals, and initializing a filter weight vector of a normalized least mean square algorithm; according to the historical previous frame audio driving signal and the intermediate state electroencephalogram signal, carrying out normalized least mean square iteration processing until all discrete time steps are traversed, and outputting a final electroencephalogram signal; The method comprises the steps of extracting audio data before a current time step in an audio driving signal of a previous historical frame and constructing a second reference vector during each iteration, carrying out convolution operation on the second reference vector by using a current filter weight vector, estimating to obtain an audio interference noise estimation signal of the current time step, filtering the audio interference noise estimation signal from an intermediate electroencephalogram signal of the current time step to obtain an electroencephalogram signal of the current time step, calculating signal energy of the second reference vector, and updating to obtain a filter weight vector of a next round by using the signal energy, the electroencephalogram signal of the current time step and the second reference vector according to a normalized minimum mean square criterion.
  6. 6. The method for processing hearing aid signals based on electroencephalogram detection according to claim 1, wherein performing neural decoding matching analysis to obtain a target sound source probability value according to the independent candidate sound source stream and the electroencephalogram signal comprises: performing gamma-pass filtering on the independent candidate sound source stream and extracting an amplitude envelope compressed by a power law to obtain auditory envelope characteristics simulating the characteristics of a human ear cochlea; band-pass filtering is carried out on the electroencephalogram signals, and nerve oscillation characteristics covering a nerve entrainment interval and an attention control interval are extracted; inputting the auditory envelope characteristic and the neural oscillation characteristic into a pre-trained convolutional neural network model, calculating matching probability in a preset time window, and outputting a target sound source probability value indicating the degree of attention of a user to a specific speaker.
  7. 7. The method for processing hearing aid signals based on electroencephalogram detection according to claim 1, wherein the performing spatial enhancement beamforming processing according to the target sound source probability value and the microphone array signal to obtain an enhanced audio signal and outputting the enhanced audio signal comprises: comparing the target sound source probability value with a preset high-low threshold value, judging the duration time to obtain a target sound source identifier, and calculating a smooth gain coefficient according to the target sound source identifier and a preset smoothing factor; according to the smooth gain coefficient, the target sound source identification and the microphone array signal, constructing a multichannel wiener filter optimization objective function comprising a noise reduction item and a space constraint item, and solving to obtain an optimal weight vector; performing linear filtering operation on the microphone array signals by using the optimal weight vector to obtain a space enhancement component, and acquiring a reference channel signal in the microphone array as a background component; And according to the smooth gain coefficient, the space enhancement component and the background component, performing weighted synthesis operation to obtain a final enhanced audio signal and outputting the final enhanced audio signal.
  8. 8. The hearing aid signal processing system based on the electroencephalogram detection is characterized by comprising a data acquisition module, a pre-processing module and a result output module; the data acquisition module is used for synchronously acquiring original brain electrical signals, microphone array signals, inertial measurement data and historical previous frame audio driving signals of a user through a hearing aid; the pre-processing module is used for performing blind source separation on the microphone array signals to obtain independent candidate sound source streams, and performing motion artifact cancellation processing and audio interference cancellation processing on the original electroencephalogram signals according to the inertia measurement data and the historical previous frame of audio driving signals to obtain electroencephalogram signals; And the result output module is used for performing neural decoding matching analysis according to the independent candidate sound source flow and the electroencephalogram signal to obtain a target sound source probability value, performing space enhancement beam forming processing according to the target sound source probability value and the microphone array signal to obtain an enhancement audio signal and outputting the enhancement audio signal.
  9. 9. A computer program product comprising a computer program or instructions which, when executed, implements a hearing aid signal processing method based on electroencephalogram detection as claimed in any one of claims 1 to 7.
  10. 10. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements a hearing aid signal processing method based on electroencephalogram detection according to any one of claims 1 to 7.

Description

Hearing aid signal processing method, system, medium and product based on electroencephalogram detection Technical Field The invention belongs to the field of hearing aid signal processing, and relates to a hearing aid signal processing method, system, medium and product based on brain electrical detection. Background As the population ages and hearing loss increases, hearing aids have been widely used as a primary intervention to enhance speech by digital signal processing. Meanwhile, the academic world proposes an auditory attention decoding technology, and aims to judge a speaker focused in a complex acoustic environment by decoding an electroencephalogram signal of a user, so as to realize intelligent sound selection. Existing high-end hearing aid products typically use an array of microphones to pick up sound, in combination with beamforming and noise reduction algorithms to enhance speech, and some attempt to introduce auditory attention decoding techniques to automatically switch programs. However, when the technology is truly landed on an in-ear hearing aid product, a high-intensity alternating magnetic field and mechanical vibration generated by a receiver during operation can be directly coupled to an electroencephalogram electrode which is closely attached to an auditory canal, so that serious micro-tone potential and motion artifacts are caused, the acquired electroencephalogram signals are annihilated by noise and completely fail, in addition, the traditional nerve-steering beam forming technology mostly adopts hard switching logic, namely, a non-target sound source is completely restrained, the processing mode breaks down spatial clues of binaural hearing, a user loses the azimuth sense of the sound source although hearing the content, the cognitive load of the brain is increased, and stable closed-loop control is difficult to maintain. Disclosure of Invention The application provides a hearing aid signal processing method, a system, a medium and a product based on electroencephalogram detection, which can solve the problems of electroencephalogram signal distortion and hearing space sense loss caused by physical interference and algorithm limitation when the traditional hearing aid is used for introducing electroencephalogram control. To achieve the above object, in a first aspect, the present invention provides a hearing aid signal processing method based on electroencephalogram detection, including: synchronously acquiring an original brain electrical signal, a microphone array signal, inertial measurement data and a historical previous frame of audio driving signal of a user through a hearing aid; performing blind source separation on the microphone array signals to obtain independent candidate sound source streams, and performing motion artifact cancellation and audio interference cancellation on the original brain electrical signals according to the inertia measurement data and the historical previous frame of audio driving signals to obtain brain electrical signals; and performing neural decoding matching analysis according to the independent candidate sound source flow and the electroencephalogram signal to obtain a target sound source probability value, performing space enhancement beam forming processing according to the target sound source probability value and the microphone array signal to obtain an enhancement audio signal and outputting the enhancement audio signal. Compared with the prior art, the embodiment of the application has the following beneficial effects that the original electroencephalogram signal, the microphone array signal, the inertial measurement data and the audio driving signal of the previous frame are synchronously acquired through the hearing aid, so as to construct a multi-mode data perception foundation; the method comprises the steps of carrying out blind source separation on microphone array signals to obtain independent candidate sound source streams, deconstructing a mixed sound field under the condition that prior sound source position information is not needed, providing independent candidate objects for attention decoding, carrying out motion artifact cancellation processing and audio interference cancellation processing on original brain electrical signals in sequence according to inertia measurement data and historical previous frame audio driving signals, wherein the historical previous frame audio driving signals are used as references, based on physical delay of sound waves transmitted from receivers to auditory canal electrodes and calculation time-consuming causality of digital signal processing, enabling a system to accurately predict and offset electromagnetic and micro-sound interference generated at the current moment by using known 'past' output signals, guaranteeing feasibility of real-time noise reduction, combining with reference of inertia measurement data, effectively cutting off electromagnetic/mechanical interference of the recei