A sound processing method, system, device and storage medium based on hearing aid

Abstract

The present invention provides a sound processing method, system, device and storage medium based on hearing aids, the method includes large-scale collection of sound signals of the hearing aid use environment, including background noise and speech signals; preprocessing the sound signals, processing outliers and missing values, and scaling them to a uniform scale, dividing them into training sets and test sets; using an improved gradient descent optimization algorithm to adjust and optimize the parameters of the standard adaptive filtering model, the parameters include filter order, maximum number of iterations, initial step size and regularization coefficient; inputting the test set into the optimized algorithm to achieve optimized processing of the sound signal. The present invention can improve speech clarity and background noise suppression capabilities, and improve the use effect and user experience of hearing aids in complex environments.

Description

Technical Field

[0001] The present invention relates to the technical field of hearing aids, and more specifically, to a sound processing method, system, device and storage medium based on a hearing aid. Background Art

[0002] In the field of hearing aid technology, sound signal processing is a key link in improving the user experience of the hearing impaired. Traditional hearing aid sound processing methods mainly rely on simple gain adjustment and fixed filtering technology. Although they can amplify the sound to a certain extent, in complex environments, such as when the background noise is strong, the clarity and intelligibility of the speech will be seriously affected. In addition, traditional methods are often unable to perform adaptive optimization according to the hearing conditions and usage environments of different users, resulting in unsatisfactory sound processing effects. In recent years, with the development of digital signal processing technology, adaptive filtering algorithms have gradually been introduced into hearing aid sound processing to achieve dynamic suppression of background noise and enhancement of speech signals. However, the existing adaptive filtering algorithms still have some limitations in practical applications, such as insufficient flexibility in parameter adjustment, inability to quickly adapt to environmental changes, and the problem of slow convergence or unstable performance during the optimization process.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: the existing hearing aid sound processing method has poor speech enhancement effect in complex environments and cannot effectively distinguish speech signals from background noise; the parameter optimization process of the adaptive filtering algorithm lacks an efficient adjustment mechanism, resulting in unstable algorithm performance and slow convergence speed, which is difficult to meet the real-time processing requirements of hearing aids. Summary of the invention

[0004] The present invention provides a hearing aid-based sound processing method, system, device and storage medium.

[0005] In a first aspect of the present invention, a hearing aid-based sound processing method is provided, comprising:

[0006] Step 1: Large-scale collection and recording of sound signals in the hearing aid usage environment, wherein the sound signals include background noise and speech signals;

[0007] Step 2: Processing abnormal values ​​and missing values ​​in the sound signal, scaling the sound signal to a uniform scale; dividing the processed sound signal into a training set and a test set;

[0008] Step 3: The training set is used to train an enhanced adaptive filtering algorithm. The enhanced adaptive filtering algorithm is specifically: using an improved gradient descent optimization algorithm to adjust and optimize the parameters of the standard adaptive filtering model. The parameters are used in the adaptive filtering model to obtain an enhanced adaptive filtering algorithm, wherein the parameters include: filter order , maximum number of iterations , initial step length And the regularization coefficient , the improved gradient descent optimization algorithm includes: designing a performance stability component and a convergence speed adjustment component to improve the update step size of the gradient descent optimization algorithm;

[0009] Step 4: Input the test set of sound signals of the hearing aid usage environment into the trained enhanced adaptive filtering algorithm to achieve optimized processing of the hearing aid sound signals.

[0010] Furthermore, in the step one, the sound signal has a total of six variable characteristics, including background noise and speech signals; wherein the background noise includes three variable characteristics, specifically including the white noise intensity, wind noise intensity and mechanical noise intensity in the environment, and the speech signal includes three variable characteristics, specifically including the frequency range of the speech, the loudness of the speech and the clarity of the speech; the sound signal is collected and recorded on a large scale, wherein the recording method is to record in the form of 6 different category variable values ​​per group.

[0011] Furthermore, in step 2, abnormal values ​​and missing values ​​in the sound signal are processed. When processing abnormal values, a specific method for finding the abnormal values ​​is as follows: in the sound signal of the hearing aid use environment, for the first Data points ,calculate To the nearest The distance between the hearing aid and the ambient sound signal data point is calculated based on the distance The local similarity of , the specific mathematical model of local similarity is:

[0012]

[0013] In formula (1), for The local similarity of for Around The most recent hearing aid use environment sound signal data points, for To The distance to the nearest hearing aid use environment sound signal data point, is the total number of recent hearing aid use environment sound signal data points, is the attenuation coefficient of distance; Using ambient sound signals for hearing aids data points;

[0014] Among them, according to calculate If the abnormality value is greater than 1, then It is an outlier, otherwise it is not. The mathematical model of the outlier value is:

[0015]

[0016] In formula (2), for The abnormality value of is the average of the local similarities of all data points.

[0017] Furthermore, in the missing value processing of step 2, the specific method of finding the missing value during the missing value processing is: dividing the hearing aid use environment sound signal data set into clusters, including , among which Cluster It is composed of a set of hearing aid use environment sound signal data points, and the first point of the hearing aid use environment sound signal is calculated using formula (3). The uncertainty score of each data point;

[0018]

[0019] In formula (3), for The uncertainty score, is the dimension of the data point, is the mean vector of the data set, is the transpose symbol, is the covariance matrix, which represents the correlation between data sets in different dimensions;

[0020] Data points with uncertainty scores higher than 0.5 are considered missing values.

[0021] Furthermore, the improved gradient descent optimization algorithm is specifically: updating the step size Improvements are made, and the improved update step size includes: performance stability component and convergence speed adjustment component , the mathematical model is:

[0022]

[0023] In formula (4), is the current iteration number;

[0024] Establishing the performance stability component , adjust the update step size by performance improvement. If the performance is significantly improved, shorten the update step size to quickly use favorable exploration directions. The performance improvement is the fitness change of individuals in the population of the improved gradient descent optimization algorithm. The mathematical model is:

[0025]

[0026] In formula (5), and are the minimum and maximum values ​​of the update step size, respectively. For the The average fitness value of the population in the iteration, For the The minimum fitness value in the iteration population, For the The minimum fitness value in the iteration population; adjust the update step size according to the volatility of performance, and increase the update frequency when the volatility is high to stabilize the behavior of the improved gradient descent optimization algorithm; where performance is the fitness value of the individuals in the improved gradient descent optimization algorithm population; the mathematical model is:

[0027]

[0028] In formula (6), is the volatility adjustment coefficient, For continuous The standard deviation of the improvement in population performance over the iterations, For continuous The mean improvement of the population performance over iterations.

[0029] Furthermore, the improved gradient descent optimization algorithm is used to adjust and optimize the parameters of the standard adaptive filtering model, and the specific steps are:

[0030] S31, the filter order of the standard adaptive filtering model , maximum number of iterations , initial step length And the regularization coefficient Encoded as a space vector, the space vector establishes a position mapping with the individual position of the improved gradient descent optimization algorithm, where the problem dimension of the improved gradient descent optimization algorithm is is 4;

[0031] S32, initialize the parameters of the improved gradient descent optimization algorithm including the maximum number of iterations , population size , Problem Dimension , search space upper bound And the search space lower bound , feedback threshold And the minimum and maximum values ​​of the update step size;

[0032] S33, Individual positions of the population of the randomly initialized improved gradient descent optimization algorithm , use formula (7) to check the individual position, if it exceeds the boundary Then the individual position is set on the boundary;

[0033]

[0034] S34, determine the current number of iterations Is it satisfied? , if satisfied, the individual position corresponding to the global minimum fitness value is output and parsed as the filter order , maximum number of iterations , initial step length And the regularization coefficient The spatial vector is recorded as the optimal parameter of the standard adaptive filtering model; otherwise, step S35 is executed;

[0035] S35, calculate the updated step size after the current iteration improvement value and current iteration number The cumulative number of iterations since the last feedback ,judge Is it greater than If so, calculate the feedback result value; otherwise, execute S37; the mathematical model of the feedback result is:

[0036]

[0037] S36, if the feedback result Greater than the balance threshold , then the local development phase of the improved gradient descent optimization algorithm is carried out, taking the current global best individual position as the center, and updating the new position of the individual; otherwise, the global exploration phase of the improved gradient descent optimization algorithm is carried out, taking the current position of the individual as the center, and introducing the Gaussian perturbation strategy to update the new position of the individual;

[0038] S37, using the objective function to calculate the individual position fitness value of the current iteratively improved gradient descent optimization algorithm population, The minimum fitness value in the iterative population is recorded as , and with The minimum fitness value in the iteration population is compared, and the smaller fitness value of the two is taken as the current optimal fitness value; the objective function The mathematical model is:

[0039]

[0040] In formula (9), For the The true value of the hearing aid sound signal, For the The optimal processing value of the hearing aid sound signal, Use the amount of ambient sound signals for hearing aids;

[0041] S38, current iteration number implement , return to step S34.

[0042] Furthermore, the Gaussian perturbation strategy is introduced to update the new position of the individual as shown in formula (10):

[0043] In formula (10), The new positions of the individuals for the improved gradient descent optimization algorithm, The best individual position for the improved gradient descent optimization algorithm, is the standard Gaussian distribution function, is the intensity of the Gaussian perturbation, A random number between 0 and 1.

[0044] In a second aspect of the present invention, there is provided a hearing aid-based sound processing system, comprising:

[0045] A sound signal acquisition module, used to collect and record the sound signals of the hearing aid usage environment on a large scale, wherein the sound signals include background noise and speech signals;

[0046] The data preprocessing module is used to process abnormal values ​​and missing values ​​in the collected sound signals and scale the sound signals to a uniform scale; the processed sound signals are divided into a training set and a test set;

[0047] The enhanced adaptive filtering algorithm training module is used to train the enhanced adaptive filtering algorithm using the training set. The enhanced adaptive filtering algorithm adjusts and optimizes the parameters of the standard adaptive filtering model through an improved gradient descent optimization algorithm. The parameters include:

[0048] Filter order N , maximum number of iterations T , initial step length μ And the regularization coefficient λ, the improved gradient descent optimization algorithm includes designing a performance stability component and a convergence speed adjustment component to improve the update step size of the gradient descent optimization algorithm;

[0049] The sound signal optimization processing module is used to input the test set into the trained enhanced adaptive filtering algorithm to achieve optimized processing of the hearing aid sound signal.

[0050] In a third aspect of the present invention, an electronic device is provided, comprising: at least one processor, a memory and an input-output unit; wherein the memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute any one of the methods described in the first aspect.

[0051] In a fourth aspect of the present invention, a computer-readable storage medium is provided, which includes instructions, and when the instructions are executed on a computer, the computer executes any one of the methods in the first aspect.

[0052] The above-mentioned embodiments according to the present invention have at least the following beneficial effects: The sound processing method based on hearing aids described in the present invention can effectively improve the speech clarity and intelligibility of hearing aids in complex environments. Through the enhanced adaptive filtering algorithm, it is possible to dynamically distinguish between speech signals and background noise, achieve accurate suppression of background noise, and enhance the quality of speech signals, thereby providing users with a clearer and more natural auditory experience. In addition, the method can also significantly improve the stability and efficiency of sound processing. The improved gradient descent optimization algorithm can flexibly adjust the filter parameters, accelerate the convergence speed of the algorithm, and ensure the stability of the performance, so that the hearing aid can quickly adapt to different usage environments and user needs, further improving the practicality and user experience of the hearing aid. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which:

[0054] Figure 1 A schematic flow chart of a sound processing method based on a hearing aid provided in one embodiment of the present invention;

[0055] Figure 2 A structural diagram of a hearing aid-based sound processing system provided by an embodiment of the present invention;

[0056] Figure 3 The schematic diagram schematically shows the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0058] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, device, apparatus, method or computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0059] It should be noted that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning.

[0060] Reference below Figure 1 , Figure 1 FIG. 1 is a flow chart of a sound processing method based on a hearing aid provided by an embodiment of the present invention. Figure 1 As shown, a hearing aid-based sound processing method 100 includes:

[0061] Step 1: Large-scale collection and recording of sound signals in the hearing aid usage environment, wherein the sound signals include background noise and speech signals;

[0062] Step 2: Processing abnormal values ​​and missing values ​​in the sound signal, scaling the sound signal to a uniform scale; dividing the processed sound signal into a training set and a test set;

[0063] Step 3: The training set is used to train an enhanced adaptive filtering algorithm. The enhanced adaptive filtering algorithm is specifically: using an improved gradient descent optimization algorithm to adjust and optimize the parameters of the standard adaptive filtering model. The parameters are used in the adaptive filtering model to obtain an enhanced adaptive filtering algorithm, wherein the parameters include: filter order , maximum number of iterations , initial step length And the regularization coefficient , the improved gradient descent optimization algorithm includes: designing a performance stability component and a convergence speed adjustment component to improve the update step size of the gradient descent optimization algorithm;

[0064] Step 4: Input the test set of sound signals of the hearing aid usage environment into the trained enhanced adaptive filtering algorithm to achieve optimized processing of the hearing aid sound signals.

[0065] It should be noted that the present invention proposes a sound processing method based on hearing aids, the core of which is to optimize the sound signal of the hearing aid by using an enhanced adaptive filtering algorithm. This method collects sound signals of the hearing aid use environment on a large scale, including background noise and speech signals, to provide a rich data basis for subsequent signal processing. In the signal processing process, the collected sound signal is first preprocessed to remove outliers and missing values, and scaled to a uniform scale to better adapt to subsequent algorithm processing. Subsequently, the parameters of the standard adaptive filtering model are adjusted and optimized by an improved gradient descent optimization algorithm, thereby achieving accurate optimization processing of the sound signal. The enhanced adaptive filtering algorithm here is a signal processing technology based on dynamic parameter adjustment, which can adjust the filtering parameters in real time according to the characteristics of environmental noise and speech signals to achieve the best speech enhancement effect. The improved gradient descent optimization algorithm is an optimization algorithm that dynamically adjusts the update step size in the gradient descent process by designing performance stability components and convergence speed adjustment components to improve the convergence speed and stability of the algorithm.

[0066] Specifically, the acquisition of sound signals is the basic step of this method. The collected sound signals include background noise and speech signals. The background noise may come from white noise, wind noise or mechanical noise in the environment, while the speech signal focuses on its frequency range, loudness and clarity. In the preprocessing stage, the processing of outliers is achieved by calculating the local similarity of data points. If the outlier value of a data point is greater than 1, it is determined to be an outlier and processed. The processing of missing values ​​is determined by calculating the uncertainty score of the data point. When the uncertainty score is higher than 0.5, the data point is determined to be a missing value. In addition, the scaling of the sound signal is to convert the collected data into a unified scale in order to better adapt to the subsequent algorithm processing. In the optimization algorithm, parameter setting is a key link. For example, parameters such as the filter order N, the maximum number of iterations T, the initial step size μ and the regularization coefficient λ need to be reasonably configured according to the specific application scenario. The optimization adjustment of these parameters can ensure the adaptability and efficiency of the algorithm in different environments.

[0067] Preferably, in order to further improve the effect of sound processing, the sampling frequency and sampling time can be increased in the signal acquisition stage to obtain more comprehensive sound signal data. In the preprocessing stage, more advanced algorithms can be used to process outliers and missing values, such as combining machine learning methods to classify and eliminate outliers, or using interpolation algorithms to fill missing values ​​more accurately. In the optimization algorithm, more performance adjustment mechanisms can be introduced, such as dynamically adjusting the weights of the performance stability component and the convergence speed adjustment component to better balance the stability and convergence speed of the algorithm. In addition, the value range of the filter order N can also be adjusted according to actual needs, such as appropriately increasing the filter order in a high noise environment to enhance noise suppression capabilities. These refinements and alternatives can further enhance the flexibility and effectiveness of the present invention in practical applications.

[0068] In some embodiments, in step one, the sound signal has six variable characteristics, including background noise and speech signal; wherein, the background noise includes three variable characteristics, specifically including the white noise intensity, wind noise intensity and mechanical noise intensity in the environment, and the speech signal includes three variable characteristics, specifically including the frequency range of speech, the loudness of speech and the clarity of speech; the sound signal is collected and recorded on a large scale, wherein the recording method is to record according to each group of 6 different category variable values.

[0069] It should be noted that the sound signal collection stage mentioned in the present invention involves six variable features, which include the features of background noise and speech signals. The three variable features of background noise are white noise intensity, wind noise intensity and mechanical noise intensity, which are used to describe different types of interference noise in the environment. The three variable features of speech signals include the frequency range, loudness and clarity of speech, which are used to characterize the quality and comprehensibility of speech signals. By recording the sound signal in the form of six different category variable values ​​per group, the sound characteristics in the hearing aid use environment can be fully captured, providing a rich data basis for subsequent sound processing. This multivariate feature collection method can more accurately reflect the sound environment in the actual use scenario, thereby providing strong support for optimizing the sound processing effect of hearing aids.

[0070] Specifically, the intensity of white noise in background noise refers to the level of random noise uniformly distributed in the environment, which is usually quantified in decibels (dB). The intensity of wind noise mainly reflects the noise intensity caused by airflow, which is particularly significant in outdoor environments, and its intensity is also measured in decibels. The intensity of mechanical noise refers to the noise level generated by mechanical equipment or other mechanical vibrations, which is also expressed in decibels. The frequency range of speech signals is usually measured in Hertz (Hz), indicating the lowest and highest frequencies of speech signals; loudness is measured in decibels, reflecting the intensity of speech; clarity is a relatively subjective indicator that can be quantified by speech recognition rate or other related algorithms. When collecting sound signals, a high-precision microphone array can be used, combined with signal processing technology, to record these variable features separately. For example, the intensity of white noise can be obtained through spectrum analysis, the intensity of wind noise can be measured with the assistance of a wind speed sensor, and the characteristics of speech signals can be extracted and quantified by speech signal processing algorithms.

[0071] Preferably, in order to further improve the accuracy and reliability of sound signal acquisition, multi-sensor fusion technology can be introduced in the acquisition process. For example, by combining multiple directional microphones, different noise sources can be collected respectively, so as to more accurately separate background noise and voice signals. In addition, adaptive noise cancellation technology can be used to remove part of the background noise in real time during the acquisition stage to improve the signal-to-noise ratio of the voice signal. In terms of recording variable features, timestamps and spatial location information can be added to better track the changing trend and source direction of the sound signal. For example, the frequency range of the voice signal can be further subdivided into multiple frequency bands for analysis to more accurately capture the characteristic changes of the voice. For the recording of background noise, an environmental noise classification algorithm can be introduced to automatically identify different types of noise sources and classify and record them according to their characteristics, thereby providing more detailed information for subsequent sound processing.

[0072] In some embodiments, the step 2 of processing abnormal values ​​and missing values ​​in the sound signal, wherein when processing abnormal values, a specific method for discovering the abnormal values ​​is: in the hearing aid use environment sound signal, for the first Data points ,calculate To the nearest The distance between the hearing aid and the ambient sound signal data point is calculated based on the distance The local similarity of , the specific mathematical model of local similarity is:

[0073]

[0074] In formula (1), for The local similarity of for Around The most recent hearing aid use environment sound signal data points, for To The distance to the nearest hearing aid use environment sound signal data point, is the total number of recent hearing aid use environment sound signal data points, is the attenuation coefficient of distance; Using ambient sound signals for hearing aids data points;

[0075] Among them, according to calculate If the abnormality value is greater than 1, then It is an outlier, otherwise it is not. The mathematical model of the outlier value is:

[0076]

[0077] In formula (2), for The abnormality value of is the average of the local similarities of all data points.

[0078] It should be noted that the present invention pays special attention to the detection and processing of outliers in the process of sound signal processing. Outliers refer to sound signal data points caused by sudden noise or other atypical factors in the hearing aid use environment. These data points may seriously deviate from the normal range, thereby affecting the accuracy and stability of subsequent sound processing. In order to detect these outliers, the present invention adopts a detection method based on local similarity. By calculating the distance between each data point and the nearest surrounding data point and constructing a local similarity model based on these distances, abnormal data points can be effectively identified. The calculation of the outlier value is based on the ratio of local similarity to overall average similarity. When the ratio is greater than 1, the data point is determined to be an outlier. This method can effectively filter out abnormal signals that may interfere with sound processing and ensure the quality and reliability of input data.

[0079] Specifically, the core of outlier detection lies in the calculation of local similarity. The local similarity model evaluates the similarity between the target data point and the nearest data points around it to evaluate its similarity with the surrounding environment. The distance attenuation coefficient in the formula determines the degree of influence of distance on similarity, and can usually be adjusted according to the actual application scenario. For example, in an environment with more complex noise, the attenuation coefficient can be appropriately increased to enhance the sensitivity to local similarity. The calculation of the outlier value is based on the ratio of local similarity to the average similarity of all data points. When the ratio is greater than 1, it means that the data point is very different from the surrounding environment and is judged as an outlier. In actual applications, the threshold of outlier detection can be dynamically adjusted according to the noise level of the hearing aid use environment and the characteristics of the speech signal to adapt to different scenario requirements.

[0080] Preferably, in order to further improve the accuracy and adaptability of outlier detection, a weighting mechanism can be introduced in the local similarity calculation. For example, different weights can be assigned to different data points according to their reliability or importance, so as to more accurately reflect the similarity between data points. In addition, the outliers can be classified and identified in combination with a machine learning algorithm, and the characteristic patterns of outliers can be automatically learned through a training model, thereby improving the efficiency and accuracy of detection. In actual operation, the parameters of outlier detection can also be dynamically adjusted according to changes in environmental noise, such as appropriately lowering the threshold of the outlier value in a high-noise environment to reduce the possibility of misjudgment. At the same time, contextual information, such as time series features or spatial location information, can be introduced to assist in determining whether a data point is an outlier, thereby further improving the robustness of outlier detection.

[0081] In some embodiments, in the missing value processing of step 2, the specific method of finding the missing value during the missing value processing is: dividing the hearing aid use environment sound signal data set into clusters, including , among which Cluster It is composed of a set of hearing aid use environment sound signal data points, and the first point of the hearing aid use environment sound signal is calculated using formula (3). The uncertainty score of each data point;

[0082]

[0083] In formula (3), for The uncertainty score, is the dimension of the data point, is the mean vector of the data set, is the transpose symbol, is the covariance matrix, which represents the correlation between data sets in different dimensions;

[0084] Data points with uncertainty scores higher than 0.5 are considered missing values.

[0085] It should be noted that the present invention pays special attention to the detection and processing of missing values ​​in the process of sound signal processing. Missing values ​​refer to the situation where the sound signal data points are lost or unreliable due to sensor failure, signal interference or other reasons in the hearing aid use environment. In order to detect these missing values, the present invention adopts a detection method based on uncertainty score. By dividing the sound signal data set into several clusters and calculating the uncertainty score of each data point, the missing values ​​can be effectively identified. The uncertainty score reflects the reliability of the data point in the data set. When its value is higher than 0.5, the data point is judged to be a missing value. This method can effectively identify data points that may affect the integrity of sound processing, thereby providing a basis for subsequent data repair and optimization processing.

[0086] Specifically, the core of missing value detection lies in the calculation of uncertainty scores. The uncertainty score is calculated by a formula that takes into account the dimensions of the data point, the mean vector of the data set, and the covariance matrix. Among them, the covariance matrix is ​​used to describe the correlation between different dimensions of the data set, while the mean vector reflects the overall distribution characteristics of the data set. In the formula, the calculation of the uncertainty score is based on the difference between the data point and the mean of the data set and the inverse matrix of the covariance matrix. When the uncertainty score is higher than 0.5, it means that the data point is significantly different from the overall data set, and is thus judged as a missing value. In practical applications, the threshold of the uncertainty score can be dynamically adjusted according to the complexity of the hearing aid use environment and the reliability of data collection. For example, in an environment with low noise levels, the threshold can be appropriately increased to reduce the possibility of misjudgment.

[0087] Preferably, in order to further improve the accuracy and adaptability of missing value detection, a weighting mechanism can be introduced in the calculation of the uncertainty score. For example, different weights are assigned to different data points according to their importance or reliability, so as to more accurately reflect the uncertainty of the data points. In addition, the missing values ​​can be predicted and repaired in combination with machine learning algorithms. The relationship between data points is automatically learned through training models to fill in missing values ​​and improve data integrity. In actual operation, time series analysis methods can also be introduced to use the continuity of data points in time to assist in determining whether a data point is a missing value. For example, if the uncertainty scores of multiple consecutive data points are high, it can be further confirmed that these data points are missing values, and interpolation or other repair methods can be used to process them.

[0088] In some embodiments, the improved gradient descent optimization algorithm is specifically: updating the step size Improvements are made, and the improved update step size includes: performance stability component and convergence speed adjustment component , the mathematical model is:

[0089]

[0090] In formula (4), is the current iteration number;

[0091] Establishing the performance stability component , adjust the update step size by performance improvement. If the performance is significantly improved, shorten the update step size to quickly use favorable exploration directions. The performance improvement is the fitness change of individuals in the population of the improved gradient descent optimization algorithm. The mathematical model is:

[0092]

[0093] In formula (5), and are the minimum and maximum values ​​of the update step size, respectively. For the The average fitness value of the population in the iteration, For the The minimum fitness value in the iteration population, For the The minimum fitness value in the iteration population; adjust the update step size according to the volatility of performance, and increase the update frequency when the volatility is high to stabilize the behavior of the improved gradient descent optimization algorithm; where performance is the fitness value of the individuals in the improved gradient descent optimization algorithm population; the mathematical model is:

[0094]

[0095] In formula (6), is the volatility adjustment coefficient, For continuous The standard deviation of the improvement in population performance over the iterations, For continuous The mean improvement of the population performance over iterations.

[0096] It should be noted that the improved gradient descent optimization algorithm mentioned in the present invention is a key technology for realizing an enhanced adaptive filtering algorithm. The algorithm improves the update step size of the gradient descent optimization algorithm by designing a performance stability component and a convergence speed adjustment component, thereby improving the optimization efficiency and stability of the algorithm. The performance stability component adjusts the update step size through fitness changes to ensure that the algorithm can quickly utilize favorable exploration directions when performance is significantly improved; the convergence speed adjustment component adjusts the update frequency according to performance volatility to stabilize the behavior of the algorithm. This improved gradient descent optimization algorithm can effectively balance the convergence speed and stability of the algorithm, making it more suitable for dynamic optimization needs in hearing aid sound signal processing.

[0097] Specifically, in the improved gradient descent optimization algorithm, the update step size consists of a performance stability component and a convergence speed adjustment component. The calculation of the performance stability component is based on the fitness change of the individuals in the population, and the degree of performance improvement is evaluated by comparing the minimum fitness value of the current iteration with that of the previous iteration. The fitness value in the formula reflects the optimization effect of the algorithm in the current iteration, and the improvement in fitness determines the adjustment direction and amplitude of the update step size. The convergence speed adjustment component evaluates the performance volatility by calculating the standard deviation and mean of the performance improvement of the population over multiple consecutive iterations. The higher the volatility, the greater the adjustment amplitude of the update step size. In practical applications, parameters such as the minimum and maximum update step sizes, the volatility adjustment coefficient, etc. need to be set according to specific problems. For example, the minimum update step size can be set to 0.01, the maximum update step size to 0.1, and the volatility adjustment coefficient can be dynamically adjusted according to the convergence of the algorithm.

[0098] Preferably, in order to further improve the performance of the improved gradient descent optimization algorithm, a dynamic adjustment mechanism can be introduced in the calculation of the performance stability component. For example, the weight of the fitness change is dynamically adjusted according to the convergence progress of the algorithm, so that the algorithm can quickly explore in the early stage and can be finely optimized in the later stage. In addition, a variety of performance volatility evaluation indicators, such as variance, range, etc., can be introduced to more comprehensively reflect the volatility of the algorithm. In the calculation of the convergence speed adjustment component, trend analysis can be combined with historical performance data to more accurately adjust the update step size. For example, when the algorithm is detected to be oscillating, the update step size is appropriately reduced to stabilize the optimization process; when the algorithm converges slowly, the update step size is appropriately increased to accelerate the optimization. In addition, the advantages of other optimization algorithms, such as crossover and mutation operations in genetic algorithms, can be combined to further enhance the global search capability of the algorithm.

[0099] In some embodiments, the parameters of the standard adaptive filtering model are adjusted and optimized using the improved gradient descent optimization algorithm, and the specific steps are:

[0100] S31, the filter order of the standard adaptive filtering model , maximum number of iterations , initial step length And the regularization coefficient Encoded as a space vector, the space vector establishes a position mapping with the individual position of the improved gradient descent optimization algorithm, where the problem dimension of the improved gradient descent optimization algorithm is is 4;

[0101] S32, initialize the parameters of the improved gradient descent optimization algorithm including the maximum number of iterations , population size , Problem Dimension , search space upper bound And the search space lower bound , feedback threshold And the minimum and maximum values ​​of the update step size;

[0102] S33, Individual positions of the population of the randomly initialized improved gradient descent optimization algorithm , use formula (7) to check the individual position, if it exceeds the boundary Then the individual position is set on the boundary;

[0103]

[0104] S34, determine the current number of iterations Is it satisfied? , if satisfied, the individual position corresponding to the global minimum fitness value is output and parsed as the filter order , maximum number of iterations , initial step length And the regularization coefficient The spatial vector is recorded as the optimal parameter of the standard adaptive filtering model; otherwise, step S35 is executed;

[0105] S35, calculate the updated step size after the current iteration improvement value and current iteration number The cumulative number of iterations since the last feedback ,judge Is it greater than If so, calculate the feedback result value; otherwise, execute S37; the mathematical model of the feedback result is:

[0106]

[0107] S36, if the feedback result Greater than the balance threshold , then the local development phase of the improved gradient descent optimization algorithm is carried out, taking the current global best individual position as the center, and updating the new position of the individual; otherwise, the global exploration phase of the improved gradient descent optimization algorithm is carried out, taking the current position of the individual as the center, and introducing the Gaussian perturbation strategy to update the new position of the individual;

[0108] S37, using the objective function to calculate the individual position fitness value of the current iteratively improved gradient descent optimization algorithm population, The minimum fitness value in the iterative population is recorded as , and with The minimum fitness value in the iteration population is compared, and the smaller fitness value of the two is taken as the current optimal fitness value; the objective function The mathematical model is:

[0109]

[0110] In formula (9), For the The true value of the hearing aid sound signal, For the The optimal processing value of the hearing aid sound signal, Use the amount of ambient sound signals for hearing aids;

[0111] S38, current iteration number implement , return to step S34.

[0112] It should be noted that the process of adjusting and optimizing the parameters of the standard adaptive filtering model using the improved gradient descent optimization algorithm mentioned in the present invention is the core link to achieve the optimization processing of sound signals. This process encodes parameters such as the filter order, maximum number of iterations, initial step size and regularization coefficient into space vectors, and establishes a mapping relationship with the individual positions of the improved gradient descent optimization algorithm, thereby achieving dynamic optimization of the parameters. The improved gradient descent optimization algorithm dynamically adjusts the update step size through the performance stability component and the convergence speed adjustment component to ensure that it can converge quickly and maintain performance stability during the optimization process. This process provides an efficient parameter adjustment mechanism for the adaptive filtering model, enabling it to better adapt to the complex sound signals in the hearing aid usage environment.

[0113] Specifically, the implementation of the improved gradient descent optimization algorithm involves the setting of multiple key parameters. The filter order (N) determines the complexity and processing power of the filter, and is usually selected according to the frequency characteristics of the sound signal. For example, in a low-frequency noise environment, a lower filter order can be set to reduce the computational complexity; while in a high-frequency noise environment, a higher filter order is required to achieve better filtering effects. The maximum number of iterations (T) is used to control the duration of the optimization process, and is usually set to 100 to 1000 iterations according to the complexity of the actual problem. The initial step size (μ) determines the initial exploration speed of the optimization process, and is generally set between 0.01 and 0.1. The specific value needs to be adjusted according to the scale and characteristics of the problem. The regularization coefficient (λ) is used to prevent overfitting and is usually set to a small positive number, such as 0.001 to 0.1. In addition, the improved gradient descent optimization algorithm also involves the setting of parameters such as the population size, the upper and lower bounds of the search space, etc. These parameters jointly determine the optimization ability and efficiency of the algorithm.

[0114] Preferably, in order to further improve the performance of the optimization algorithm, more flexibility and adaptability can be introduced in the parameter setting and optimization process. For example, the filter order can be dynamically adjusted according to the dynamic characteristics of the sound signal, so that it can adaptively select the optimal value under different environments. In the setting of the maximum number of iterations, an adaptive termination condition can be introduced. When the performance improvement of multiple consecutive iterations is lower than a certain threshold, the optimization process is terminated in advance to improve computational efficiency. For the initial step size, an adaptive adjustment strategy can be adopted to dynamically adjust the step size according to the performance changes of the current iteration, so as to quickly explore in the early stage of optimization and make fine adjustments in the later stage of optimization. In addition, a multi-objective optimization mechanism can be introduced to consider multiple objectives such as noise suppression and speech fidelity at the same time, and optimize the filter parameters by weighing these objectives, so as to further improve the effect of hearing aid sound processing.

[0115] In some embodiments, a Gaussian perturbation strategy is introduced to update the new position of the individual as shown in formula (10):

[0116] In formula (10), The new positions of the individuals for the improved gradient descent optimization algorithm, The best individual position for the improved gradient descent optimization algorithm, is the standard Gaussian distribution function, is the intensity of the Gaussian perturbation, A random number between 0 and 1.

[0117] It should be noted that the purpose of introducing the Gaussian perturbation strategy in the present invention to update the new position of the individual is to enhance the global exploration ability of the improved gradient descent optimization algorithm. During the optimization process, the algorithm may fall into a local optimal solution and thus fail to find the global optimal solution. By introducing the Gaussian perturbation strategy, random perturbations can be introduced based on the current position of the individual, thereby increasing the probability of the algorithm jumping out of the local optimal solution and improving the global search capability. The Gaussian perturbation strategy generates random perturbations based on the standard Gaussian distribution function, and its intensity can be adjusted by parameters to ensure that the algorithm can maintain stability during the optimization process and effectively explore new solution spaces.

[0118] Specifically, the implementation of the Gaussian perturbation strategy involves the setting of key parameters. Among them, the intensity of the Gaussian perturbation (σ) is an important parameter, which determines the amplitude of the random perturbation. The intensity value is usually set according to the complexity of the problem and the optimization goal, generally between 0.01 and 0.1. For example, when the optimization goal is more complex or the search space is larger, the perturbation intensity can be appropriately increased to enhance the global exploration ability; when the optimization goal is relatively simple, the perturbation intensity can be reduced to increase the convergence speed of the algorithm. Random numbers (rand) are uniformly distributed random numbers between 0 and 1, which are used to combine with the Gaussian distribution function to generate specific perturbation values. The Gaussian perturbation strategy calculates the new position of the individual through a formula, that is, the perturbation value is added to the current best individual position, thereby updating the individual position. This strategy can not only help the algorithm escape the local optimal solution, but also improve the robustness of the algorithm to a certain extent.

[0119] Preferably, in order to further enhance the effect of the Gaussian perturbation strategy, an adaptive mechanism can be introduced in the setting of the perturbation intensity. For example, the perturbation intensity is dynamically adjusted according to the performance improvement of the current iteration. When the algorithm detects that the performance improvement is slow or falls into a local optimum, the perturbation intensity is automatically increased to enhance the global exploration capability; when the performance improvement is obvious, the perturbation intensity is appropriately reduced to speed up the convergence speed. In addition, a combination of multiple perturbation strategies can be introduced, such as combining uniformly distributed perturbations or other non-Gaussian distributed perturbations, to enrich the exploration method of the algorithm. In the process of updating the new position of an individual, a boundary constraint mechanism can be introduced to ensure that the updated individual position does not exceed the predefined search space range. For example, when the new position exceeds the boundary, it can be adjusted to the boundary or the perturbation value can be regenerated to ensure the stability and effectiveness of the algorithm.

[0120] The above-mentioned embodiments of the present invention have the following beneficial effects: The hearing aid-based sound processing method described in the present invention can realize efficient optimization processing of hearing aid sound signals. By collecting sound signals of the hearing aid usage environment on a large scale and performing preprocessing, outliers and missing values ​​can be effectively removed to ensure the accuracy and reliability of input data. The improved gradient descent optimization algorithm can flexibly adjust the parameters of the standard adaptive filtering model, such as the filter order, the maximum number of iterations, the initial step size, and the regularization coefficient, so as to achieve accurate optimization of the sound signal. In addition, the design of the performance stability component and the convergence speed adjustment component can further improve the adaptability and convergence efficiency of the algorithm, so that the hearing aid can respond quickly and provide high-quality sound processing effects in different environments.

[0121] At the same time, this method can also significantly improve the user experience and practicality of hearing aids. By applying the optimized algorithm to the test set, the clarity and intelligibility of the speech signal can be effectively enhanced, while suppressing the interference of background noise, providing a more comfortable and natural auditory experience for the hearing impaired. The introduction of the Gaussian perturbation strategy further enhances the global exploration capability of the algorithm, avoiding falling into the local optimal solution, thereby ensuring the stability and consistency of the sound processing effect. Overall, the present invention provides an efficient, flexible and reliable solution for hearing aid sound processing, which can meet the usage needs of different users in diverse environments.

[0122] like Figure 2 As shown, a hearing aid-based sound processing system 200 in some embodiments includes:

[0123] The sound signal collection module 201 is used to collect and record the sound signals of the hearing aid use environment on a large scale, wherein the sound signals include background noise and speech signals;

[0124] The data preprocessing module 202 is used to process abnormal values ​​and missing values ​​in the collected sound signals and scale the sound signals to a uniform scale; and divide the processed sound signals into a training set and a test set;

[0125] The enhanced adaptive filtering algorithm training module 203 is used to train the enhanced adaptive filtering algorithm using the training set. The enhanced adaptive filtering algorithm adjusts and optimizes the parameters of the standard adaptive filtering model through an improved gradient descent optimization algorithm. The parameters include:

[0126] Filter order N , maximum number of iterations T , initial step length μ And the regularization coefficient λ, the improved gradient descent optimization algorithm includes designing a performance stability component and a convergence speed adjustment component to improve the update step size of the gradient descent optimization algorithm;

[0127] The sound signal optimization processing module 204 is used to input the test set into the trained enhanced adaptive filtering algorithm to achieve optimized processing of the hearing aid sound signal.

[0128] It is understandable that the modules described in the hearing aid-based sound processing system 200 are similar to those described in the reference Figure 1 Therefore, the operations, features and beneficial effects described above for the sound processing method based on the hearing aid are also applicable to the sound processing system 200 based on the hearing aid and the modules contained therein, and will not be described in detail here.

[0129] Reference below Figure 3 , which shows a schematic diagram of the structure of an electronic device 300 suitable for implementing some embodiments of the present invention. The electronic devices in some embodiments of the present invention may include but are not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0130] like Figure 3 As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. In the RAM, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM, and the RAM are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0131] Typically, the following devices may be connected to the I / O interface: input devices 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 308 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 309. The communication devices 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0132] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0133] The above descriptions are only some preferred embodiments of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. A sound processing method based on a hearing aid, characterized in that: The enhanced adaptive filtering algorithm is used to optimize the processing of hearing aid sound signals. The specific steps are: Step 1: Large-scale collection and recording of sound signals in the hearing aid usage environment, wherein the sound signals include background noise and speech signals; Step 2: Processing abnormal values ​​and missing values ​​in the sound signal, scaling the sound signal to a uniform scale; dividing the processed sound signal into a training set and a test set; Step 3: The training set is used to train an enhanced adaptive filtering algorithm. The enhanced adaptive filtering algorithm is specifically: using an improved gradient descent optimization algorithm to adjust and optimize the parameters of the standard adaptive filtering model. The parameters are used in the adaptive filtering model to obtain an enhanced adaptive filtering algorithm, wherein the parameters include: filter order , maximum number of iterations , initial step length And the regularization coefficient , the improved gradient descent optimization algorithm includes: designing a performance stability component and a convergence speed adjustment component to improve the update step size of the gradient descent optimization algorithm; Step 4: Input the test set of sound signals of the hearing aid usage environment into the trained enhanced adaptive filtering algorithm to optimize the processing of the hearing aid sound signals; The step 2 is to process the abnormal values ​​and missing values ​​in the sound signal, wherein when processing the abnormal values, the specific method of finding the abnormal values ​​is: in the hearing aid use environment sound signal, for the first Data points ,calculate To the nearest The distance between the hearing aid and the ambient sound signal data point is calculated based on the distance The local similarity of , the specific mathematical model of local similarity is: In formula (1), for The local similarity of for Around The most recent hearing aid use environment sound signal data points, for To The distance to the nearest hearing aid use environment sound signal data point, is the total number of recent hearing aid use environment sound signal data points, is the attenuation coefficient of distance; Using ambient sound signals for hearing aids data points; Among them, according to calculate If the abnormality value is greater than 1, then It is an outlier, otherwise it is not. The mathematical model of the outlier value is: In formula (2), for The abnormality value of is the average of the local similarities of all data points.

2. A hearing aid-based sound processing method according to claim 1, characterized in that: In the step one, the sound signal has six variable characteristics, including background noise and speech signal; wherein the background noise includes three variable characteristics, specifically including the white noise intensity, wind noise intensity and mechanical noise intensity in the environment, and the speech signal includes three variable characteristics, specifically including the frequency range of speech, the loudness of speech and the clarity of speech; the sound signal is collected and recorded on a large scale, wherein the recording method is to record in the form of 6 different category variable values ​​per group.

3. The sound processing method based on a hearing aid according to claim 1, characterized in that: In the missing value processing of step 2, the specific method of finding the missing value during the missing value processing is: dividing the hearing aid use environment sound signal data set into clusters, including , among which Cluster It is composed of a set of hearing aid use environment sound signal data points, and the first point of the hearing aid use environment sound signal is calculated using formula (3). The uncertainty score of each data point; In formula (3), for The uncertainty score, is the dimension of the data point, is the mean vector of the data set, is the transpose symbol, is the covariance matrix, which represents the correlation between data sets in different dimensions; Data points with uncertainty scores higher than 0.5 are considered missing values.

4. The sound processing method based on a hearing aid according to claim 1, characterized in that: The improved gradient descent optimization algorithm is specifically: update step size Improvements are made, and the improved update step size includes: performance stability component and convergence speed adjustment component , the mathematical model is: In formula (4), is the current iteration number; Establishing the performance stability component , adjust the update step size by performance improvement. If the performance is significantly improved, shorten the update step size to quickly use favorable exploration directions. The performance improvement is the fitness change of individuals in the population of the improved gradient descent optimization algorithm. The mathematical model is: In formula (5), and are the minimum and maximum values ​​of the update step size, respectively. For the The average fitness value of the population in the iteration, For the The minimum fitness value in the iteration population, For the The minimum fitness value in the iteration population; adjust the update step size according to the volatility of performance, and increase the update frequency when the volatility is high to stabilize the behavior of the improved gradient descent optimization algorithm; where performance is the fitness value of the individuals in the improved gradient descent optimization algorithm population; the mathematical model is: In formula (6), is the volatility adjustment coefficient, For continuous The standard deviation of the improvement in population performance over the iterations, For continuous The mean improvement of the population performance over iterations.

5. The hearing aid-based sound processing method according to claim 1, characterized in that: The improved gradient descent optimization algorithm is used to adjust and optimize the parameters of the standard adaptive filtering model, and the specific steps are: S31, the filter order of the standard adaptive filtering model , maximum number of iterations , initial step length And the regularization coefficient Encoded as a space vector, the space vector establishes a position mapping with the individual position of the improved gradient descent optimization algorithm, where the problem dimension of the improved gradient descent optimization algorithm is is 4; S32, initialize the parameters of the improved gradient descent optimization algorithm including the maximum number of iterations , population size , Problem Dimension , search space upper bound And the search space lower bound , feedback threshold And the minimum and maximum values ​​of the update step size; S33, Individual positions of the population of the randomly initialized improved gradient descent optimization algorithm , use formula (7) to check the individual position, if it exceeds the boundary Then the individual position is set on the boundary; S34, determine the current number of iterations Is it satisfied? , if satisfied, the individual position corresponding to the global minimum fitness value is output and parsed as the filter order , maximum number of iterations , initial step length And the regularization coefficient The spatial vector is recorded as the optimal parameter of the standard adaptive filtering model; otherwise, step S35 is executed; S35, calculate the updated step size after the current iteration improvement value and current iteration number The cumulative number of iterations since the last feedback ,judge Is it greater than If so, calculate the feedback result value; otherwise, execute S37; the mathematical model of the feedback result is: S36, if the feedback result Greater than the balance threshold , then the local development phase of the improved gradient descent optimization algorithm is carried out, taking the current global best individual position as the center, and updating the new position of the individual; otherwise, the global exploration phase of the improved gradient descent optimization algorithm is carried out, taking the current position of the individual as the center, and introducing the Gaussian perturbation strategy to update the new position of the individual; S37, using the objective function to calculate the individual position fitness value of the current iteratively improved gradient descent optimization algorithm population, The minimum fitness value in the iterative population is recorded as , and with Compare the minimum fitness values ​​in the iteration population, and take the smaller fitness value as the current best fitness value; Objective Function The mathematical model is: In formula (9), For the The true value of the hearing aid sound signal, For the The optimal processing value of the hearing aid sound signal, Use the amount of ambient sound signals for hearing aids; S38, current iteration number implement , return to step S34.

6. The method according to claim 5, characterized in that The Gaussian perturbation strategy is introduced to update the new position of the individual as shown in formula (10): In formula (10), The new positions of the individuals for the improved gradient descent optimization algorithm, The best individual position for the improved gradient descent optimization algorithm, is the standard Gaussian distribution function, is the intensity of the Gaussian perturbation, A random number between 0 and 1.

7. A sound processing system based on a hearing aid, characterized in that: include: A sound signal acquisition module, used to collect and record the sound signals of the hearing aid usage environment on a large scale, wherein the sound signals include background noise and speech signals; A data preprocessing module is used to process abnormal values ​​and missing values ​​in the collected sound signals and scale the sound signals to a uniform scale; Divide the processed sound signal into a training set and a test set; The enhanced adaptive filtering algorithm training module is used to train the enhanced adaptive filtering algorithm using the training set. The enhanced adaptive filtering algorithm adjusts and optimizes the parameters of the standard adaptive filtering model through an improved gradient descent optimization algorithm. The parameters include: Filter order N , maximum number of iterations T , initial step length μ And the regularization coefficient λ , the improved gradient descent optimization algorithm includes designing a performance stability component and a convergence speed adjustment component to improve the update step size of the gradient descent optimization algorithm; The sound signal optimization processing module is used to input the test set into the trained enhanced adaptive filtering algorithm to achieve optimized processing of the hearing aid sound signal; The abnormal values ​​and missing values ​​in the sound signal are processed, wherein when the abnormal value is processed, the specific method of finding the abnormal value is: in the hearing aid use environment sound signal, for the first Data points ,calculate To the nearest The distance between the hearing aid and the ambient sound signal data point is calculated based on the distance The local similarity of , the specific mathematical model of local similarity is: In formula (1), for The local similarity of for Around The most recent hearing aid use environment sound signal data points, for To The distance to the nearest hearing aid use environment sound signal data point, is the total number of recent hearing aid use environment sound signal data points, is the attenuation coefficient of distance; Using ambient sound signals for hearing aids data points; Among them, according to calculate If the abnormality value is greater than 1, then It is an outlier, otherwise it is not. The mathematical model of the outlier value is: In formula (2), for The abnormality value of is the average of the local similarities of all data points.

8. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images