Intelligent hearing aid noise reduction method combined with environmental noise evaluation
By identifying the type of environmental noise through noise time-varying index and spectral flatness, and adaptively adjusting the noise reduction parameters based on the user's hearing characteristics, the problem of insufficient adaptability of hearing aids in complex environments is solved, achieving personalized noise reduction effects and improved speech clarity.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-26
- Publication Date
- 2026-04-07
AI Technical Summary
Existing hearing aid noise reduction technologies are not adaptable enough to complex environments and cannot meet the personalized needs of users, resulting in unsatisfactory noise reduction effects and speech distortion.
The environmental noise type is identified by two-dimensional features of noise time-varying index and spectral flatness. The noise reduction parameters are adaptively adjusted by combining the environmental noise type and the user's hearing characteristics, and differentiated noise tracking and noise reduction gain processing are adopted.
It improves the accuracy of noise classification and environmental adaptability, achieves personalized noise reduction effects, and enhances the noise reduction performance and speech clarity of hearing aids in complex environments.
Smart Images
Figure CN121811906A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of hearing aid signal processing technology, and in particular to a smart hearing aid noise reduction method that incorporates environmental noise assessment. Background Technology
[0002] Hearing aids are electronic devices that help people with hearing impairments improve their hearing. Their core function is to extract and enhance speech signals from noisy environments. Noise reduction technology is one of the key factors affecting the performance of hearing aids. Existing noise reduction methods for hearing aids are generally based on noise power spectrum estimation and adaptive filtering techniques. For example, spectral subtraction estimates the noise power spectrum and subtracts noise components from the noisy signal, Wiener filtering calculates the filter gain based on the minimum mean square error criterion, and the minimum mean square error filter updates the filter coefficients through an adaptive algorithm. These methods can achieve certain noise reduction effects under laboratory conditions.
[0003] However, hearing aids face complex and varied acoustic environments in actual use, including traffic noise, wind noise, and crowd noise, among others, and the noise characteristics change with time and space. Existing noise reduction technologies suffer from the following problems when dealing with these complex environments: First, the noise reduction effect is often insufficient, resulting in residual noise or speech distortion in certain noisy environments, affecting the user's auditory comfort and speech intelligibility; second, existing hearing aids typically use preset noise reduction parameters, failing to adequately consider individual differences such as the degree of hearing loss and auditory preferences of different users, making it difficult to provide personalized noise reduction effects. Therefore, improving the adaptability of hearing aid noise reduction systems in complex environments and their ability to personalize for different users is a pressing technical problem that needs to be solved in this field. Summary of the Invention
[0004] This invention provides an intelligent hearing aid noise reduction method that incorporates environmental noise assessment, addressing the technical problems of insufficient adaptability of existing hearing aid noise reduction technologies in complex environments and difficulty in meeting users' personalized needs.
[0005] The first aspect of this invention provides a method for noise reduction in a smart hearing aid that incorporates environmental noise assessment, comprising: Audio signals are collected through the microphone of the hearing aid, and the audio signals are processed by frame segmentation and converted to the frequency domain to obtain frequency domain signals. Noise features are extracted from frequency domain signals, the noise time-varying index and spectral flatness are calculated, and the type of environmental noise is identified based on the noise time-varying index and spectral flatness. The noise power spectrum is estimated based on the frequency domain signal, the noise tracking parameters are determined according to the type of environmental noise, and the noise power spectrum is updated using the noise tracking parameters. Calculate the signal-to-noise ratio parameter corresponding to the noise power spectrum, determine the noise reduction intensity factor according to the type of environmental noise and the user's hearing characteristics, and calculate the noise reduction gain based on the signal-to-noise ratio parameter and the noise reduction intensity factor. The noise reduction gain is used to denoise the frequency domain signal and then convert it to the time domain to output the processed audio signal.
[0006] Optionally, the audio signal is framed and converted to the frequency domain to obtain a frequency domain signal, including: The hearing aid's microphone collects ambient sound to obtain analog audio signals; Analog audio signals are sampled and quantized using an analog-to-digital converter to obtain digital audio signals; The digital audio signal is preprocessed to obtain the preprocessed digital audio signal; The preprocessed digital audio signal is overlapped and framed to obtain multiple time-domain audio frames; Windowing is performed on each temporal audio frame using a window function to obtain windowed audio frames; Perform a Fast Fourier Transform on each windowed audio frame to obtain the frequency domain signal.
[0007] Optionally, the noise time-varying index and spectral flatness are calculated, including: Detect speech activity segments in the frequency domain signal, calculate the spectral change rate between adjacent frames in non-speech activity segments, and obtain the noise time-varying index by statistically analyzing the spectral change rate. The frequency domain signal is divided into multiple sub-bands and the energy of each sub-band is calculated. The spectral flatness is then calculated based on the energy of each sub-band.
[0008] Optionally, the type of ambient noise can be identified based on the noise time-varying index and spectral flatness, including: The time-varying characteristics of environmental noise are determined by the noise time-varying index, and the spectral characteristics of environmental noise are determined by the spectral flatness. Based on time-varying and spectral characteristics, environmental noise can be classified into steady-state broadband noise, steady-state narrowband noise, or non-steady-state noise. The noise classification results of multiple consecutive frames are processed in a time sequence to output a stable environmental noise type.
[0009] Optionally, noise tracking parameters are determined based on the type of ambient noise, and the noise power spectrum is updated using the noise tracking parameters, including: The method for updating the noise power spectrum is determined based on the type of environmental noise. When the ambient noise type is steady-state broadband noise, time smoothing is performed on the noise power spectrum of the current frame and the historical noise power spectrum, and frequency domain smoothing is performed on adjacent frequency points. When the ambient noise type is steady-state narrowband noise, the noise power spectrum of the current frame and the historical noise power spectrum are smoothed over time, and each frequency point is updated independently. When the ambient noise type is non-steady-state noise, increase the update weight of the noise power spectrum of the current frame, and adjust the frequency domain smoothing range according to the spectral structure of the current frame. The inter-frame energy change of the noise power spectrum is detected, and the initial estimation is re-executed when the energy change exceeds the preset change amount. Output the updated noise power spectrum.
[0010] Optionally, the noise reduction gain is calculated based on the signal-to-noise ratio parameter and the noise reduction intensity factor, including: Calculate the posterior signal-to-noise ratio at each frequency point based on the power spectrum and noise power spectrum of the frequency domain signal; Determine the basic noise reduction intensity factor based on the type of environmental noise; Acquire user audiogram data, calculate hearing adaptation coefficient based on the degree of hearing loss at each frequency point, and use the hearing adaptation coefficient to adjust the basic noise reduction intensity factor in a frequency-adaptive manner to obtain a personalized noise reduction intensity factor. Identify key speech frequency bands in the frequency domain signal and set noise reduction protection limits for these key speech frequency bands; Based on the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the previous frame, a decision-guided method is used to estimate the prior signal-to-noise ratio of the current frame. The prior signal-to-noise ratio is adjusted by using a personalized noise reduction intensity factor and then limited by combining the noise reduction protection upper limit to obtain the adjusted prior signal-to-noise ratio. Based on the adjusted prior and posterior signal-to-noise ratios, calculate the noise reduction gain at each frequency point.
[0011] Optionally, noise reduction processing of the frequency domain signal is performed using noise reduction gain, including: The amplitude spectrum of the frequency domain signal is weighted frequency-by-frequency using the noise reduction gain to obtain the noise-reduced amplitude spectrum; The denoised amplitude spectrum is combined with the original phase spectrum to obtain the denoised frequency domain signal; Perform a fast inverse Fourier transform on the denoised frequency domain signal to obtain the denoised time domain audio frame. Adjacent denoised time-domain audio frames are overlapped and added together to obtain a continuous time-domain audio signal, which is then output as a time-domain audio signal.
[0012] A second aspect of the present invention provides an intelligent hearing aid noise reduction system that incorporates environmental noise assessment, comprising: The signal acquisition module is used to acquire audio signals through the microphone of the hearing aid, perform frame-by-frame processing on the audio signals, and convert them to the frequency domain to obtain frequency domain signals; The noise classification module is used to determine the noise time-varying index and spectral flatness based on the frequency domain signal, and to identify the type of environmental noise based on the noise time-varying index and spectral flatness. The noise estimation module is used to estimate the noise power spectrum in the non-speech activity segment of the frequency domain signal, determine the noise tracking parameters according to the type of environmental noise, and update the noise power spectrum using the noise tracking parameters. The gain calculation module is used to calculate the signal-to-noise ratio parameter corresponding to the noise power spectrum, determine the noise reduction intensity factor according to the type of environmental noise and the user's hearing characteristics, and calculate the noise reduction gain based on the signal-to-noise ratio parameter and the noise reduction intensity factor. The noise reduction module is used to perform noise reduction processing on the frequency domain signal using noise reduction gain, and then convert it to the time domain to output the processed audio signal.
[0013] A computer device provided in a third aspect of the present invention includes a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the intelligent hearing aid noise reduction method incorporating environmental noise assessment as described above.
[0014] The fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, performs the steps as described in any of the above claims.
[0015] The beneficial effects of this invention are as follows: It employs a dual-dimensional feature of noise time-varying index and spectral flatness to identify environmental noise types, which can more accurately reflect the inherent characteristics of noise, improving the accuracy of noise classification and environmental adaptability. Based on the noise type, parameters are adaptively adjusted in both the noise estimation and noise reduction stages, forming a dual adaptive mechanism that better balances noise reduction depth and speech fidelity, achieving good noise reduction effects in various noise environments. Furthermore, by combining environmental noise type and user hearing characteristics to determine noise reduction intensity, it achieves an organic combination of environmental adaptation and personalized processing. Compared to a uniform noise reduction strategy, it can provide more precise noise reduction solutions for users with different degrees of hearing loss, improving the user experience of hearing aids and speech clarity. Attached Figure Description
[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is a flowchart illustrating a method for noise reduction in an intelligent hearing aid that incorporates environmental noise assessment, as provided in an embodiment of the present invention.
[0018] Figure 2 This is a flowchart illustrating the environmental noise type identification process of an intelligent hearing aid noise reduction method that incorporates environmental noise assessment, as provided in an embodiment of the present invention.
[0019] Figure 3 This is a flowchart illustrating the noise power spectrum estimation and updating process of a smart hearing aid noise reduction method that incorporates environmental noise assessment, as provided in an embodiment of the present invention.
[0020] Figure 4 This is a structural block diagram of an intelligent hearing aid noise reduction system that incorporates environmental noise assessment, provided as an embodiment of the present invention. Detailed Implementation
[0021] This invention provides an intelligent hearing aid noise reduction method that incorporates environmental noise assessment, addressing the technical problems of insufficient adaptability of existing hearing aid noise reduction technologies in complex environments and difficulty in meeting users' personalized needs.
[0022] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0023] like Figure 1 As shown, the present invention provides a method for noise reduction in intelligent hearing aids that incorporates environmental noise assessment, comprising: S1: Audio signals are acquired through the microphone of the hearing aid, the audio signals are processed by frame segmentation and converted to the frequency domain to obtain frequency domain signals.
[0024] Specifically, S1 includes: S1.1: Acquire ambient sound through the microphone of the hearing aid to obtain analog audio signals.
[0025] S1.2: The analog audio signal is sampled and quantized by an analog-to-digital converter to obtain a digital audio signal.
[0026] In this embodiment, the sampling rate of the analog-to-digital converter is not less than 16kHz, and the quantization bit depth is not less than 16bit.
[0027] S1.3: Preprocess the digital audio signal to obtain the preprocessed digital audio signal.
[0028] The preprocessing includes at least one of the following: DC component removal, automatic gain control, and notch filtering. Specifically, DC component removal is used to eliminate the DC bias introduced by the analog-to-digital converter, automatic gain control is used to adjust the signal amplitude to a suitable range, and notch filtering is used to remove power frequency interference.
[0029] S1.4: Overlap the preprocessed digital audio signal into multiple time-domain audio frames.
[0030] Furthermore, the frame length is determined according to the time resolution requirements of speech spectrum analysis, and an appropriate frame length is selected to balance the spectrum resolution and time resolution; the frame shift is determined according to the inter-frame continuity requirements, and the frame shift is less than the frame length; starting from the beginning position of the preprocessed digital audio signal, signal segments of length equal to the frame length are extracted sequentially according to the frame shift to obtain multiple time-domain audio frames.
[0031] S1.5: Window each time-domain audio frame using a window function to obtain windowed audio frames.
[0032] Windowing is used to reduce spectral leakage in the Fast Fourier Transform. Optionally, the window function can be a Hamming window or a Hanning window.
[0033] S1.6: Perform a Fast Fourier Transform on each windowed audio frame to obtain the frequency domain signal.
[0034] In this embodiment, the frequency domain signal includes the amplitude spectrum and phase spectrum at each frequency point.
[0035] S2: Extract noise features from the frequency domain signal, calculate the noise time-varying index and spectral flatness, and identify the type of environmental noise based on the noise time-varying index and spectral flatness.
[0036] Specifically, the flowchart for identifying environmental noise types is as follows: Figure 2 As shown, it includes: S2.1: Detect speech activity segments in the frequency domain signal, calculate the spectral change rate between adjacent frames in non-speech activity segments, and obtain the noise time-varying index by statistically analyzing the spectral change rate.
[0037] In this embodiment, the energy of each frequency point of the frequency domain signal is calculated, and the frame energy is obtained by summing the energies of each frequency point; the spectral kurtosis of the frequency domain signal is calculated, and the current frame is determined to be a speech activity segment based on the frame energy and spectral kurtosis; in frames determined to be non-speech activity segments, the normalized difference between the spectrum of the current frame and the spectrum of the previous frame is calculated to obtain the inter-frame spectral change rate; outlier removal is performed on the inter-frame spectral change rate of multiple consecutive frames, and the mean and variance of the inter-frame spectral change rate after outlier removal are statistically analyzed; the noise time-varying index is calculated based on the mean and variance.
[0038] The spectral kurtosis is defined as the fourth-order normalized moment of the frequency domain power spectrum, and its calculation formula is as follows:
[0039] in, For spectral kurtosis, This represents the total number of frequency points in the current frame. The power spectrum at the f-th frequency point, The mean of the power spectrum of the current frame. This represents the standard deviation of the power spectrum. Due to the formant structure, the spectral energy of a speech segment is concentrated at specific frequencies, typically exhibiting... The peak distribution characteristics; the spectral energy distribution of the noise segment is more uniform, closer to a Gaussian distribution, exhibiting the following characteristics. .
[0040] Furthermore, the threshold for determining speech activity segments is adaptively set. During the initial calibration phase, frame energy and spectral kurtosis are statistically analyzed. The energy threshold is the high quantile of the frame energy during the calibration phase plus a margin, and the kurtosis threshold is the median of the spectral kurtosis during the calibration phase. When both frame energy and spectral kurtosis exceed their respective thresholds, a speech activity segment is identified. During operation, the thresholds for non-speech segment samples are periodically recalculated to adapt to changes in environmental noise.
[0041] In the outlier removal process, the median absolute deviation method is used to process the inter-frame spectral change rate sequence of consecutive frames. First, the median and median absolute deviation of the sequence are calculated, and then outlier samples that deviate too much from the median are removed. For the remaining samples after outlier removal, the mean and variance are calculated. The noise time-varying index is determined based on the ratio of variance to mean, which reflects the temporal stability of environmental noise.
[0042] S2.2: Divide the frequency domain signal into multiple sub-bands and calculate the energy of each sub-band, and calculate the spectral flatness based on the energy of each sub-band.
[0043] Preferably, the effective frequency range for spectrum analysis is determined, and the frequency domain signal is divided into multiple sub-bands within the effective frequency range; the sum of the energy of all frequency points in each sub-band is calculated to obtain the energy of each sub-band; in non-speech activity segments, the geometric mean and arithmetic mean are calculated based on the energy of each sub-band, and the ratio of the geometric mean to the arithmetic mean is calculated to obtain the spectrum flatness; the spectrum flatness of multiple consecutive frames is smoothed over time to output a stable spectrum flatness.
[0044] S2.3: Determine the time-varying characteristics of environmental noise based on the noise time-varying index, and determine the spectral characteristics of environmental noise based on the spectral flatness.
[0045] In practice, the noise time-varying index is compared with a preset time-varying threshold. When the noise time-varying index is less than the preset threshold, the time-varying characteristic of the environmental noise is determined to be steady-state; when the noise time-varying index is greater than or equal to the preset threshold, the time-varying characteristic of the environmental noise is determined to be non-steady-state. The spectral flatness is compared with a preset spectral threshold. When the spectral flatness is greater than or equal to the preset threshold, the spectral characteristic of the environmental noise is determined to be broadband; when the spectral flatness is less than the preset threshold, the spectral characteristic of the environmental noise is determined to be narrowband.
[0046] The preset time-varying threshold and preset spectral threshold are determined based on the statistical characteristics of a large number of noise samples. Typical steady-state and non-steady-state noise samples are collected, and the statistical distribution of their noise time-varying exponents is calculated. The discrimination boundary between the two noise distributions is selected as the preset time-varying threshold. The preset spectral threshold is determined using a similar method based on the statistical distribution of the spectral flatness of broadband and narrowband noise samples.
[0047] S2.4: Based on time-varying and spectral characteristics, environmental noise is classified into steady-state broadband noise, steady-state narrowband noise, or non-steady-state noise.
[0048] Furthermore, when the time-varying characteristics are steady-state and the spectral characteristics are broadband, the environmental noise is classified as steady-state broadband noise; when the time-varying characteristics are steady-state and the spectral characteristics are narrowband, the environmental noise is classified as steady-state narrowband noise; when the time-varying characteristics are non-steady-state, the environmental noise is classified as non-steady-state noise.
[0049] S2.5: Performs temporal smoothing on the noise classification results of multiple consecutive frames to output a stable environmental noise type.
[0050] It should be noted that the temporal smoothing process calculates the frequency of occurrence of each noise type within a preset time window and outputs the noise type with the highest frequency as the stable environmental noise type. The preset time window length is set to 0.2s to 1s; the noise type determination result is updated once per frame.
[0051] Through the above-described S2 step, this invention achieves accurate identification of environmental noise types. Specifically, by extracting noise features from non-speech activity segments, interference from speech signals in noise characteristic determination is avoided, improving the accuracy of noise classification. By calculating two orthogonal features—the noise time-varying exponent and spectral flatness—a comprehensive characterization of noise's time and frequency domain characteristics is achieved. By classifying noise into three types—steady-state broadband noise, steady-state narrowband noise, and non-steady-state noise—a basis is provided for subsequent adoption of differentiated noise reduction strategies for different noise types, thereby improving the noise reduction effect of hearing aids and the user experience in complex acoustic environments.
[0052] S3: Estimate the noise power spectrum based on the frequency domain signal, determine the noise tracking parameters according to the type of environmental noise, and update the noise power spectrum using the noise tracking parameters.
[0053] Specifically, the flowchart for noise power spectrum estimation and updating is as follows: Figure 3 As shown, it includes: S3.1: Extract the minimum spectral value of the frequency domain signal in the non-speech activity segment, and estimate the noise power spectrum based on the minimum spectral value.
[0054] In one embodiment, the energy of each frequency point of the frequency domain signal is extracted in the non-voice activity segment; within a preset statistical window, the energy of each frequency point is tracked for minimum value; the minimum value of each frequency point is smoothed to obtain the initial noise power spectrum; the validity of the initial noise power spectrum is detected, and if the validity does not meet the requirements, the minimum value tracking is re-executed.
[0055] The validity of the initial noise power spectrum is determined through a stability test, which evaluates the fluctuation characteristics of the noise power spectrum in the time and frequency domains. The preset range is determined based on the statistical distribution characteristics of typical noise; fluctuations exceeding this range are considered invalid. If the validity test fails, the minimum value tracking is re-executed after adjusting the statistical window length or the minimum value tracking parameter.
[0056] S3.2: Determine the update method for the noise power spectrum based on the type of environmental noise.
[0057] Specifically, the time smoothing coefficient and frequency domain processing method are determined based on the type of environmental noise. The time smoothing coefficient is determined according to the noise change rate; for steady-state noise, a larger coefficient close to 1 is used to obtain a stable estimate, while for non-steady-state noise, a smaller coefficient is used to quickly track noise changes.
[0058] Furthermore, when the ambient noise type is steady-state broadband noise, a larger time smoothing coefficient (e.g., 0.85~0.95) is set, and a frequency domain processing method of joint smoothing of adjacent frequency points is selected; when the ambient noise type is steady-state narrowband noise, a larger time smoothing coefficient (e.g., 0.8~0.95) is set, and a frequency domain processing method of independent updating of each frequency point is selected; when the ambient noise type is non-steady-state noise, a smaller time smoothing coefficient (e.g., 0.3~0.6) is set, and an adaptive frequency domain processing method is selected.
[0059] S3.3: When the ambient noise type is steady-state broadband noise, perform time smoothing on the noise power spectrum of the current frame and the historical noise power spectrum, and perform frequency domain smoothing on adjacent frequency points.
[0060] In practice, the power spectrum estimate of the current frame and the historical noise power spectrum are weighted and averaged for each frequency point to obtain the time-smoothed noise power spectrum; among them, the historical noise power spectrum has a larger weight and the current frame has a smaller weight, so as to achieve slow smooth tracking.
[0061] Furthermore, the time-smoothed noise power spectrum is smoothed in the frequency domain, and the noise power spectrum of each frequency point is weighted and averaged with the noise power spectrum of its adjacent frequency points. Taking advantage of the similar energy distribution of steady-state broadband noise at adjacent frequency points, the stability of noise power spectrum estimation is further improved.
[0062] S3.4: When the ambient noise type is steady-state narrowband noise, perform time smoothing on the noise power spectrum of the current frame and the historical noise power spectrum, and update each frequency point independently.
[0063] In another embodiment, the power spectrum estimate of the current frame is weighted and averaged with the historical noise power spectrum for each frequency point, and the same time smoothing coefficient as the steady-state broadband noise is used to achieve slow smooth tracking.
[0064] It should be noted that independent updating of each frequency point means that frequency domain smoothing is not performed on adjacent frequencies, thus maintaining the independence of noise power spectrum estimation for each frequency point. The reason for this approach is that the energy of steady-state narrowband noise is concentrated in a specific frequency band. If adjacent frequencies are smoothed, the noise energy will diffuse into the clean frequency band, reducing the accuracy of noise estimation.
[0065] S3.5: When the ambient noise type is non-steady-state noise, increase the update weight of the noise power spectrum of the current frame and adjust the frequency domain smoothing range according to the spectral structure of the current frame.
[0066] Specifically, for each frequency point, the power spectrum estimate of the current frame is weighted and averaged with the historical noise power spectrum. Unlike steady-state noise, the weight of the current frame is increased to achieve timely tracking of rapidly changing noise.
[0067] Furthermore, the frequency domain smoothing range is adjusted based on the energy difference between each frequency point and its adjacent frequencies. When the energy difference between a frequency point and its adjacent frequencies is small, it indicates a smooth local spectrum, and the smoothing range is expanded to jointly process that frequency point and several of its adjacent frequencies. When the energy difference is large, it indicates a spectral abrupt change or narrowband interference, and the smoothing range is narrowed to process only that frequency point and its immediate neighbors. Through this frequency-by-frequency adaptive adjustment, the noise power spectrum update can both quickly respond to sudden noise and maintain the accuracy of the estimation.
[0068] S3.6: Detect the inter-frame energy change of the noise power spectrum. When the energy change exceeds the preset change amount, re-execute the initial estimation.
[0069] Further, the ratio of the total energy of the noise power spectrum of the current frame to the total energy of the noise power spectrum of the previous frame is calculated; when the energy ratio is greater than the preset upper limit threshold or less than the preset lower limit threshold, it is determined to be a sudden change in the noise power spectrum, triggering the re-execution of the initial estimation of S3.1.
[0070] The preset upper threshold and preset lower threshold are determined based on the statistical characteristics of noise energy changes under different acoustic environments. In one embodiment, the preset upper threshold is 2.0~3.0, and the preset lower threshold is 0.3~0.5.
[0071] S3.7: Output the updated noise power spectrum.
[0072] Through step S3 described above, this invention employs a differentiated update strategy based on the type of environmental noise, thereby improving the accuracy of noise power spectrum estimation. Specifically, independent frequency-point updates prevent the spread of narrowband noise, adaptive frequency domain smoothing enables rapid tracking of non-steady-state noise, and energy mutation detection and rapid re-initialization mechanisms shorten the adaptation time during scene switching. These technical effects enable hearing aids to accurately estimate the noise power spectrum in complex acoustic environments, providing a reliable foundation for subsequent adaptive noise reduction processing.
[0073] S4: Calculate the signal-to-noise ratio parameter corresponding to the noise power spectrum, determine the noise reduction intensity factor based on the type of environmental noise and the user's hearing characteristics, and calculate the noise reduction gain based on the signal-to-noise ratio parameter and the noise reduction intensity factor.
[0074] The signal-to-noise ratio (SNR) parameter includes the prior SNR and the posterior SNR.
[0075] Specifically, S4 includes: S4.1: Calculate the posterior signal-to-noise ratio at each frequency point based on the power spectrum and noise power spectrum of the frequency domain signal.
[0076] In one embodiment, the modulus square operation is performed on the frequency domain signal of the current frame to obtain the noisy speech power spectrum at each frequency point; the noise power spectrum is obtained, and numerical stability processing is performed on the noise power spectrum. When the noise power spectrum is less than a preset noise floor threshold, it is set as the noise floor threshold; the posterior signal-to-noise ratio at each frequency point is calculated based on the noisy speech power spectrum and the noise power spectrum; and the posterior signal-to-noise ratio is range-limited.
[0077] The noise floor threshold is determined based on the background noise level of the hearing aid microphone. It is calculated by measuring the microphone's noise power in a silent environment and adding an appropriate margin as the noise floor threshold to prevent an excessively low noise power spectrum from causing abnormal signal-to-noise ratio calculations.
[0078] S4.2: Determine the basic noise reduction intensity factor based on the type of environmental noise.
[0079] Specifically, the environmental noise type is determined; when the environmental noise type is steady-state broadband noise, a larger base noise reduction factor is set (e.g., 0.75~0.85); when the environmental noise type is steady-state narrowband noise, the frequency points corresponding to the narrowband frequency band are identified, and a larger base noise reduction factor is set for the narrowband frequency points (e.g., 0.8~0.9), while a medium base noise reduction factor is set for the non-narrowband frequency points (e.g., 0.5~0.65); when the environmental noise type is non-steady-state noise, a smaller base noise reduction factor is set (e.g., 0.25~0.4).
[0080] S4.3: Obtain the user's audiogram data, calculate the hearing adaptation coefficient based on the degree of hearing loss corresponding to each frequency point, and use the hearing adaptation coefficient to adjust the basic noise reduction intensity factor in a frequency-adaptive manner to obtain a personalized noise reduction intensity factor.
[0081] Preferably, the user's audiogram data is acquired, which includes hearing thresholds corresponding to each frequency point; the degree of hearing loss is calculated based on the hearing thresholds at each frequency point; a hearing adaptation coefficient is calculated based on the degree of hearing loss, and the degree of hearing loss is converted into a hearing adaptation coefficient through an inverse mapping relationship, with the frequency point corresponding to the greater the degree of hearing loss having a smaller hearing adaptation coefficient; the basic noise reduction intensity factor is multiplied by the hearing adaptation coefficient to obtain the personalized noise reduction intensity factor for each frequency point; the personalized noise reduction intensity factor is then subjected to frequency domain smoothing to avoid excessive differences between adjacent frequency points.
[0082] S4.4: Identify key speech frequency bands in the frequency domain signal and set noise reduction protection limits for key speech frequency bands.
[0083] In another embodiment, the frequency range of key speech frequency bands is determined, including the main speech frequency band and the high-frequency clear auxiliary audio band; a noise reduction protection upper limit is set for the key speech frequency bands to avoid excessive noise reduction from damaging speech clarity; the noise reduction protection upper limit is used to limit the maximum attenuation of subsequent noise reduction gain.
[0084] S4.5: Based on the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the previous frame, the decision-guided method is used to estimate the prior signal-to-noise ratio of the current frame.
[0085] In specific implementation, the posterior signal-to-noise ratio (SNR) of the current frame, the prior SNR of the previous frame, the noise reduction gain of the previous frame, and the posterior SNR of the previous frame are obtained; a smoothing estimation term is calculated based on the data of the previous frame; a direct estimation term is calculated based on the posterior SNR of the current frame; a weighted average is performed on the smoothing estimation term and the direct estimation term to obtain the prior SNR of the current frame; and a lower limit protection process is applied to the prior SNR.
[0086] In one embodiment, when processing the first frame, the posterior signal-to-noise ratio of the current frame is used as the initial value of the prior signal-to-noise ratio.
[0087] S4.6: Adjust the prior signal-to-noise ratio using a personalized noise reduction intensity factor and limit it in conjunction with the noise reduction protection upper limit to obtain the adjusted prior signal-to-noise ratio.
[0088] In practical applications, the prior signal-to-noise ratio (SNR), personalized noise reduction intensity factor, and noise reduction protection upper limit are obtained. The prior SNR is multiplied by the personalized noise reduction intensity factor to obtain the initially adjusted prior SNR. It is then determined whether each frequency point is located in the critical speech frequency band. For frequency points located in the critical speech frequency band, the initially adjusted prior SNR is compared with the noise reduction protection upper limit, and the smaller of the two values is taken as the adjusted prior SNR. For frequency points not located in the critical speech frequency band, the initially adjusted prior SNR is used as the adjusted prior SNR.
[0089] S4.7: Calculate the noise reduction gain at each frequency point based on the adjusted prior signal-to-noise ratio and posterior signal-to-noise ratio.
[0090] Accordingly, the adjusted prior signal-to-noise ratio (SNR) and posterior signal-to-noise ratio (SNR) are obtained; the theoretical noise reduction gain at each frequency point is calculated based on the adjusted prior and posterior SNR; a lower limit is imposed on the theoretical noise reduction gain to prevent the useful signal from being excessively suppressed and to avoid excessive attenuation; the noise reduction gain is smoothed in the frequency domain to reduce abrupt changes in gain between adjacent frequency points; and the noise reduction gain is smoothed in the time domain to avoid inter-frame gain jumps.
[0091] It should be noted that a lower limit is set for the noise reduction gain to ensure that a moderate amount of ambient sound information is preserved while reducing noise, so as to avoid users experiencing an unnatural sense of isolation.
[0092] Through the above S4 steps, this invention sets differentiated noise reduction intensity factors according to the type of environmental noise, improving the adaptability of noise reduction; it adaptively adjusts the noise reduction intensity at each frequency point based on the user's audiogram data, allowing frequencies with severe hearing loss to retain more signal energy, thus achieving personalized noise reduction; and by setting a noise reduction protection upper limit for key speech frequency bands, it protects speech intelligibility while reducing noise. These technical effects enable the hearing aid to dynamically adjust noise reduction parameters according to different noise environments and user hearing characteristics, improving noise reduction performance and speech intelligibility.
[0093] S5: Use noise reduction gain to perform noise reduction processing on the frequency domain signal and convert it to the time domain to output the processed audio signal.
[0094] Specifically, the amplitude spectrum of the frequency domain signal is weighted point-by-point using the noise reduction gain to obtain the noise-reduced amplitude spectrum: obtain the frequency domain signal of the current frame and the corresponding noise reduction gain; calculate the amplitude spectrum of the frequency domain signal; multiply the noise reduction gain with the amplitude spectrum point-by-point to obtain the noise-reduced amplitude spectrum.
[0095] Furthermore, the amplitude spectrum after noise reduction is synthesized with the original phase spectrum to obtain the noise-reduced frequency domain signal; the noise-reduced frequency domain signal is subjected to inverse fast Fourier transform to obtain the noise-reduced time domain audio frame; adjacent noise-reduced time domain audio frames are overlapped and added to obtain a continuous time domain audio signal, and the time domain audio signal is output.
[0096] The beneficial effects of this invention are as follows: It employs a dual-dimensional feature of noise time-varying index and spectral flatness to identify environmental noise types, which can more accurately reflect the inherent characteristics of noise, improving the accuracy of noise classification and environmental adaptability. Based on the noise type, parameters are adaptively adjusted in both the noise estimation and noise reduction stages, forming a dual adaptive mechanism that better balances noise reduction depth and speech fidelity, achieving good noise reduction effects in various noise environments. Furthermore, by combining environmental noise type and user hearing characteristics to determine noise reduction intensity, it achieves an organic combination of environmental adaptation and personalized processing. Compared to a uniform noise reduction strategy, it can provide more precise noise reduction solutions for users with different degrees of hearing loss, improving the user experience of hearing aids and speech clarity.
[0097] like Figure 4 As shown, the present invention provides an intelligent hearing aid noise reduction system that incorporates environmental noise assessment, comprising: The signal acquisition module is used to acquire audio signals through the microphone of the hearing aid, perform frame-by-frame processing on the audio signals, and convert them to the frequency domain to obtain frequency domain signals.
[0098] The noise classification module is used to determine the noise time-varying index and spectral flatness based on the frequency domain signal, and to identify the type of environmental noise based on the noise time-varying index and spectral flatness.
[0099] The noise estimation module is used to estimate the noise power spectrum in the non-speech activity segment of the frequency domain signal, determine the noise tracking parameters according to the type of ambient noise, and update the noise power spectrum using the noise tracking parameters.
[0100] The gain calculation module is used to calculate the signal-to-noise ratio parameter corresponding to the noise power spectrum, determine the noise reduction intensity factor based on the type of environmental noise and the user's hearing characteristics, and calculate the noise reduction gain based on the signal-to-noise ratio parameter and the noise reduction intensity factor.
[0101] The noise reduction module is used to perform noise reduction processing on the frequency domain signal using noise reduction gain, and then convert it to the time domain to output the processed audio signal.
[0102] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system and modules described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0103] This invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program; when the computer program is executed by the processor, the processor performs the steps of the intelligent hearing aid noise reduction method incorporating environmental noise assessment as described in any of the above embodiments.
[0104] This invention also provides a computer-readable storage medium storing a computer program / instructions thereon, which, when executed by a processor, implement the steps of the intelligent hearing aid noise reduction method incorporating environmental noise assessment as described in any of the above embodiments.
[0105] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.
[0106] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0107] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for noise reduction in intelligent hearing aids that incorporates environmental noise assessment, characterized in that, include: Audio signals are acquired through the microphone of the hearing aid, and the audio signals are processed by frame segmentation and converted to the frequency domain to obtain frequency domain signals. Noise features are extracted from the frequency domain signal, the noise time-varying index and spectral flatness are calculated, and the type of environmental noise is identified based on the noise time-varying index and spectral flatness. Based on the frequency domain signal, the noise power spectrum is estimated; based on the environmental noise type, the noise tracking parameters are determined; and the noise power spectrum is updated using the noise tracking parameters. Calculate the signal-to-noise ratio parameter corresponding to the noise power spectrum, determine the noise reduction intensity factor according to the environmental noise type and user hearing characteristics, and calculate the noise reduction gain based on the signal-to-noise ratio parameter and the noise reduction intensity factor; The noise reduction gain is used to perform noise reduction processing on the frequency domain signal and then converted to the time domain to output the processed audio signal.
2. The intelligent hearing aid noise reduction method combining environmental noise assessment according to claim 1, characterized in that, The step of performing frame segmentation processing on the audio signal and converting it to the frequency domain to obtain a frequency domain signal includes: The hearing aid's microphone collects ambient sound to obtain analog audio signals; The analog audio signal is sampled and quantized by an analog-to-digital converter to obtain a digital audio signal; The digital audio signal is preprocessed to obtain a preprocessed digital audio signal; The preprocessed digital audio signal is overlapped and framed to obtain multiple time-domain audio frames; Windowing is performed on each temporal audio frame using a window function to obtain windowed audio frames; Perform a Fast Fourier Transform on each of the windowed audio frames to obtain the frequency domain signal.
3. The intelligent hearing aid noise reduction method combining environmental noise assessment according to claim 1, characterized in that, The calculation of the noise time-varying index and spectral flatness includes: Detect speech activity segments in the frequency domain signal, calculate the spectral change rate between adjacent frames in non-speech activity segments, and obtain the noise time-varying index by statistically analyzing the spectral change rate. The frequency domain signal is divided into multiple sub-bands and the energy of each sub-band is calculated. The spectral flatness is then calculated based on the energy of each sub-band.
4. The intelligent hearing aid noise reduction method combining environmental noise assessment according to claim 1, characterized in that, The step of identifying the type of environmental noise based on the noise time-varying index and spectral flatness includes: The time-varying characteristics of environmental noise are determined based on the noise time-varying index, and the spectral characteristics of environmental noise are determined based on the spectral flatness. Based on the time-varying and spectral characteristics, environmental noise is classified into steady-state broadband noise, steady-state narrowband noise, or non-steady-state noise. The noise classification results of multiple consecutive frames are processed in a time sequence to output a stable environmental noise type.
5. The intelligent hearing aid noise reduction method combining environmental noise assessment according to claim 1, characterized in that, The step of determining noise tracking parameters based on the environmental noise type and updating the noise power spectrum using the noise tracking parameters includes: The method for updating the noise power spectrum is determined based on the type of environmental noise. When the ambient noise type is steady-state broadband noise, time smoothing is performed on the noise power spectrum of the current frame and the historical noise power spectrum, and frequency domain smoothing is performed on adjacent frequency points. When the ambient noise type is steady-state narrowband noise, the noise power spectrum of the current frame and the historical noise power spectrum are smoothed over time, and each frequency point is updated independently. When the ambient noise type is non-steady-state noise, increase the update weight of the noise power spectrum of the current frame, and adjust the frequency domain smoothing range according to the spectral structure of the current frame. The inter-frame energy change of the noise power spectrum is detected, and the initial estimation is re-executed when the energy change exceeds the preset change amount. Output the updated noise power spectrum.
6. The intelligent hearing aid noise reduction method combining environmental noise assessment according to claim 1, characterized in that, The calculation of noise reduction gain based on the signal-to-noise ratio parameter and the noise reduction intensity factor includes: Calculate the posterior signal-to-noise ratio at each frequency point based on the power spectrum and noise power spectrum of the frequency domain signal; Determine the basic noise reduction intensity factor based on the type of environmental noise; Acquire user audiogram data, calculate hearing adaptation coefficients based on the degree of hearing loss corresponding to each frequency point, and use the hearing adaptation coefficients to perform frequency adaptive adjustment on the basic noise reduction intensity factor to obtain a personalized noise reduction intensity factor. Identify key speech frequency bands in the frequency domain signal and set a noise reduction protection limit for the key speech frequency bands; Based on the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the previous frame, the prior signal-to-noise ratio of the current frame is estimated using a decision-guided method. The prior signal-to-noise ratio is adjusted using the personalized noise reduction intensity factor and limited by the noise reduction protection upper limit to obtain the adjusted prior signal-to-noise ratio. Based on the adjusted prior signal-to-noise ratio and the posterior signal-to-noise ratio, the noise reduction gain at each frequency point is calculated.
7. The intelligent hearing aid noise reduction method combining environmental noise assessment according to claim 1, characterized in that, The noise reduction process using the noise reduction gain to denoise the frequency domain signal includes: The amplitude spectrum of the frequency domain signal is weighted frequency-by-frequency using the noise reduction gain to obtain the noise-reduced amplitude spectrum; The denoised amplitude spectrum is combined with the original phase spectrum to obtain the denoised frequency domain signal; Perform a fast inverse Fourier transform on the denoised frequency domain signal to obtain the denoised time domain audio frame. Adjacent denoised time-domain audio frames are overlapped and added together to obtain a continuous time-domain audio signal, which is then output.
8. A smart hearing aid noise reduction system that incorporates environmental noise assessment, characterized in that, include: The signal acquisition module is used to acquire audio signals through the microphone of the hearing aid, perform frame-by-frame processing on the audio signals, and convert them to the frequency domain to obtain frequency domain signals. A noise classification module is used to determine the noise time-varying index and spectral flatness based on the frequency domain signal, and to identify the type of environmental noise based on the noise time-varying index and spectral flatness. The noise estimation module is used to estimate the noise power spectrum of the non-speech activity segment of the frequency domain signal, determine noise tracking parameters according to the environmental noise type, and update the noise power spectrum using the noise tracking parameters. The gain calculation module is used to calculate the signal-to-noise ratio parameter corresponding to the noise power spectrum, determine the noise reduction intensity factor according to the environmental noise type and the user's hearing characteristics, and calculate the noise reduction gain based on the signal-to-noise ratio parameter and the noise reduction intensity factor. The noise reduction processing module is used to perform noise reduction processing on the frequency domain signal using the noise reduction gain, and convert it to the time domain to output the processed audio signal.
9. A computer device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the intelligent hearing aid noise reduction method incorporating environmental noise assessment as described in any one of claims 1-7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the steps of the intelligent hearing aid noise reduction method incorporating environmental noise assessment as described in any one of claims 1-7.