Signal processing method of single-channel hearing aid and single-channel hearing aid
By employing time-frequency transformation and mixed noise power spectral density algorithms in single-channel hearing aids to dynamically adjust gain synthesis, the problem of balancing sound quality and clarity in single-channel hearing aids under different environments is solved, achieving a low-cost and efficient auditory experience.
Patent Information
- Application Number
- CN202511617397.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-06
- Publication Date
- 2026-01-20
AI Technical Summary
Existing single-channel hearing aids cannot intelligently balance sound quality comfort and speech intelligibility in different listening scenarios, and their fixed gain synthesis method cannot adapt to changes in the real-time acoustic environment.
By employing time-frequency transformation and hybrid noise power spectral density algorithms, dynamic range compression gain and adaptive noise reduction gain are calculated by dividing virtual frequency sub-bands, and the synthesized gain is dynamically adjusted according to environmental parameters to achieve adaptive processing of spectral representation.
Achieving multi-channel processing effects under a single-channel architecture provides clear speech signals and a comfortable listening experience. It has the advantages of low cost, low power consumption, and small size, making it suitable for complex listening environments.
Smart Images
Figure CN121367871A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of hearing aids, in particular to a signal processing method of a single-channel hearing aid and a single-channel hearing aid. BACKGROUND
[0002] A hearing aid is an important auxiliary device for compensating for hearing loss. In order to accurately match the individual hearing loss curve of a user, a multi-channel digital hearing aid can provide excellent sound quality and speech intelligibility by dividing an audio signal into multiple independent frequency bands for processing. However, a multi-channel architecture usually requires a high-performance processor, a large-capacity memory, and a complex hardware design, resulting in high cost, high power consumption, and difficulty in miniaturization, which limits its popularity among a wider population.
[0003] As an alternative, a single-channel hearing aid has the advantages of low cost, low power consumption, and easy miniaturization. In order to improve the performance of a single-channel hearing aid, some existing technologies have proposed a scheme for simulating multi-channel processing under a single-channel hardware architecture. For example, some schemes analyze the acquired single-channel signal in the frequency domain, separately calculate a dynamic range compression gain for compensating for hearing loss and a noise reduction gain for improving the signal-to-noise ratio, and then apply the two gains to the signal to generate an output, such as simply multiplying or adding the two gain values. However, such existing technologies have a significant drawback: the synthesis method of the dynamic range compression gain and the noise reduction gain is usually fixed and cannot be adaptively adjusted according to the real-time changes in the acoustic environment. This static synthesis strategy cannot achieve the optimal balance of listening experience in different listening scenarios. For example, in a quiet environment, unnecessary noise reduction processing can introduce distortion and affect the naturalness of sound quality; while in a noisy environment, a fixed noise reduction intensity may not be sufficient to ensure clear and intelligible speech. Therefore, the existing technologies cannot intelligently balance sound quality comfort and speech intelligibility, making it difficult to meet the comprehensive listening experience needs of users in varying environments. SUMMARY
[0004] To this end, the present application provides a signal processing method of a single-channel hearing aid and a single-channel hearing aid, aiming to solve the technical problem in the prior art that the gain synthesis method is fixed and cannot adaptively balance sound quality comfort and speech intelligibility according to real-time changes in the acoustic environment.
[0005] To achieve the above object, the present application adopts the following technical solutions: According to a first aspect of the present application, a signal processing method of a single-channel hearing aid is provided, the method comprising: obtaining a time-domain signal frame of a single-channel audio signal, and performing time-frequency transformation on the time-domain signal frame to obtain a spectral representation of the time-domain signal frame in the frequency domain; In the frequency domain, the frequency spectrum is divided into a plurality of virtual frequency subbands, a corresponding dynamic range compression gain is calculated for each virtual frequency subband, and a hybrid noise power spectral density algorithm is used to solve the background noise of the frequency spectrum to analyze an adaptive noise reduction gain for suppressing the background noise. According to an environmental parameter of the single-channel audio signal, a synthesis manner of the dynamic range compression gain and the adaptive noise reduction gain is determined, and a comprehensive gain is generated based on the synthesis manner; The comprehensive gain is applied to the frequency spectrum representation, and a frequency-time inverse transform is performed on the updated frequency spectrum representation to reconstruct a target audio signal for output.
[0006] Further, the time-domain signal frame of the single-channel audio signal includes: An original single-channel audio signal is obtained, and data preprocessing is performed on the single-channel audio signal to obtain the time-domain signal frame; The data preprocessing includes weighting filtering, framing, and windowing operations.
[0007] Further, the frequency spectrum is divided into a plurality of virtual frequency subbands, and a corresponding dynamic range compression gain is calculated for each virtual frequency subband, including: The frequency spectrum is virtually segmented according to a preset frequency band to obtain a plurality of virtual frequency subbands; Sound pressure level calculation is performed on the virtual frequency subbands, and a preset compression gain rule matched with the sound pressure level is used to calculate a dynamic range compression gain corresponding to each virtual frequency subband; The preset compression gain rule includes: For a virtual frequency subband with a sound pressure level less than a compression start threshold, a fixed linear gain is executed; For a virtual frequency subband with a sound pressure level reaching the compression start threshold and less than a compression end threshold, a smooth transition gain based on a preset compression ratio is executed; For a virtual frequency subband with a sound pressure level reaching the compression end threshold, a fixed ratio attenuation is executed.
[0008] Further, the background noise of the frequency spectrum is solved using a hybrid noise power spectral density algorithm to analyze an adaptive noise reduction gain for suppressing the background noise, including: The target power spectrum of the background noise is estimated based on a minimum statistical noise estimation algorithm and an active speech detection algorithm; Based on the target power spectrum, an optimal modified log spectrum amplitude estimation algorithm is used to calculate an adaptive noise reduction gain.
[0009] Further, the method for estimating the target power spectrum of the background noise based on the minimum statistical noise estimation algorithm and the active speech detection algorithm comprises: Based on the minimum statistical noise estimation algorithm, the minimum power of each frequency point is tracked in a preset time window to obtain an initial power spectrum of the background noise; and based on the active speech detection algorithm, speech activity detection is performed on a current signal frame to determine whether the current signal frame contains speech components. If the current signal frame does not contain speech components, the initial power spectrum is corrected to obtain an updated target power spectrum of the background noise. If the current signal frame contains speech components, the initial power spectrum is taken as the target power spectrum of the background noise.
[0010] Further, the method for performing speech activity detection on a current signal frame based on the active speech detection algorithm to determine whether the current signal frame contains speech components comprises: An active speech detection algorithm based on multiple target features is used to calculate the posterior signal-to-noise ratio, spectral flatness and spectral entropy of the current signal frame. According to preset condition thresholds matched with the target features, it is determined whether the target features meet speech activity conditions. When the target features all meet the speech activity conditions, or when a previous signal frame of the current signal frame contains speech components and the posterior signal-to-noise ratio meets a speech activity condition, it is determined that the current signal frame contains speech components.
[0011] Further, the method for calculating an adaptive noise reduction gain based on the power spectrum of the background noise using an optimal correction log spectrum amplitude estimation algorithm comprises: The ratio of the power spectrum of noisy speech to the power spectrum of the background noise is calculated to obtain a posterior signal-to-noise ratio; and the ratio of the power spectrum of clean speech to the power spectrum of the background noise is calculated to obtain an a priori signal-to-noise ratio. The speech presence probability is determined according to the posterior signal-to-noise ratio and the a priori signal-to-noise ratio, and the optimal gain for each frequency point is dynamically adjusted based on the speech presence probability, so that speech components obtain high gain and noise components obtain low gain.
[0012] Further, the method for determining a synthesis mode of the dynamic range compression gain and the adaptive noise reduction gain according to the environmental parameters of the single-channel audio signal, and generating a comprehensive gain based on the synthesis mode comprises: compute an average signal-to-noise ratio of the single-channel audio signal, determine a synthesis weight of the dynamic range compression gain and the adaptive noise reduction gain based on a preset synthesis rule matched with the average signal-to-noise ratio, and perform weighted summation on the dynamic range compression gain and the adaptive noise reduction gain according to the synthesis weight to obtain the comprehensive gain; The preset synthesis rule comprises: when the average signal-to-noise ratio is higher than a first preset threshold, a higher weight is assigned to the dynamic range compression gain; when the average signal-to-noise ratio is lower than a second preset threshold, a higher weight is assigned to the adaptive noise reduction gain; and the first preset threshold is greater than the second preset threshold.
[0013] Further, the method further comprises: performing spectral smoothing processing and / or time smoothing processing on the adaptive noise reduction gain to suppress musical noise; The spectral smoothing processing comprises smoothing processing on gain values of adjacent frequency points, and the time smoothing processing comprises smoothing processing on gain values of the same frequency point in consecutive time frames.
[0014] According to the second aspect of the present application, a single-channel hearing aid is provided, which comprises, in sequence, a sound collecting module, an analog-to-digital conversion module, a core processing module, a digital-to-analog conversion module and a sound output module. The sound output module is configured to capture environmental sound information and convert the environmental sound information into an initial electrical signal. The analog-to-digital conversion module is configured to perform analog-to-digital conversion processing on the electrical signal to obtain an original single-channel audio signal. The core processing module is configured to perform the signal processing method of the single-channel hearing aid according to any one of the first aspect of the present application. The digital-to-analog conversion module is configured to perform digital-to-analog conversion processing on a target audio signal output by the core processing module to obtain a target electrical signal. The sound output module is configured to convert the target electrical signal into a sound wave for output.
[0015] The present application has at least the following beneficial effects: The application provides a signal processing method of a single-channel hearing aid and the single-channel hearing aid, and the method comprises the following steps: acquiring a time domain signal frame of a single-channel audio signal, and performing time-frequency conversion on the time domain signal frame to obtain a frequency spectrum representation of the time domain signal frame in a frequency domain; dividing the frequency spectrum into a plurality of virtual frequency subbands in the frequency domain, calculating a corresponding dynamic range compression gain for each virtual frequency subband; and solving background noise of the frequency spectrum by using a hybrid noise power spectral density algorithm to analyze an adaptive noise reduction gain for suppressing the background noise; determining a combination mode of the dynamic range compression gain and the adaptive noise reduction gain according to an environmental parameter of the single-channel audio signal, generating a comprehensive gain based on the combination mode; applying the comprehensive gain to the frequency spectrum representation, and performing frequency-time inverse conversion on the updated frequency spectrum representation to reconstruct a target audio signal for output. By the application, the multi-channel processing effect is realized under the single-channel architecture, the hearing aid can still provide clear speech signals and comfortable listening experience in a complex listening environment, and the application has the beneficial effects of low cost, low power consumption and small size.
[0016] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory and are not restrictive of the application. BRIEF DESCRIPTION OF DRAWINGS
[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.
[0018] Figure 1 A flowchart of a signal processing method of a single-channel hearing aid provided by an embodiment of the present application is shown; Figure 2 A structural diagram of a signal processing device of a single-channel hearing aid provided by an embodiment of the present application is shown; Figure 3 A structural diagram of a single-channel hearing aid provided by an embodiment of the present application is shown. DETAILED DESCRIPTION
[0019] The exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.
[0020] It is to be noted that, in the present document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.
[0021] The embodiment of the present application provides a signal processing method of a single-channel hearing aid, as shown in the figure, which can at least include the following steps S101-S104: Figure 1 Step S101, acquiring a time domain signal frame of a single-channel audio signal, and performing time-frequency conversion on the time domain signal frame to obtain a spectral representation of the time domain signal frame in a frequency domain.
[0022] Firstly, an original single-channel audio signal is acquired, data preprocessing is performed on the single-channel audio signal, and a time domain signal frame is obtained. The data preprocessing can include weighting filtering, framing and windowing operations.
[0023] The weighting filtering refers to enhancing high-frequency components in the audio signal through a high-frequency boosting filter, so that high-frequency speech or environmental sound is more easily heard. The transfer function of the filter can be set as: wherein, =0.97 represents an attenuation coefficient, which is used to compensate for high-frequency hearing loss.
[0024] The framing and windowing operations are pre-steps for converting continuous time domain signals into discrete frequency domain analysis. Framing cuts continuous time domain long signals into overlapping short frame segments. The framing can be performed in a 50% overlapping manner. Windowing (such as applying a Hanning window) can reduce spectral leakage at the frame edges, and provide high-quality input for subsequent FFT, power spectrum calculation and the like.
[0025] Further, after the time domain signal frame is acquired, a fast Fourier transform (FFT) is performed on each time domain signal frame to convert it from a time domain to a frequency domain, and then a spectral representation of the signal frame is obtained. For example, for a 256-point time domain signal frame, after performing a 256-point fast Fourier transform, a complex spectrum containing 129 independent frequency points (from a direct current component to a Nyquist frequency) can be obtained. The spectral representation contains amplitude and phase information of the audio signal at different frequencies, and constitutes the basis for all subsequent frequency domain processing.
[0026] In step S102, the spectrum is divided into a plurality of virtual frequency subbands in the frequency domain, and a corresponding dynamic range compression gain is calculated for each virtual frequency subband; and a background noise of the spectrum is solved by using a hybrid noise power spectral density algorithm to analyze an adaptive noise reduction gain for suppressing the background noise.
[0027] Step S102 includes two parallel core processing steps S102-1 and S102-2, and the specific embodiments of steps S102-1 and S102-2 are introduced as follows: S102-1, the spectrum is divided into a plurality of virtual frequency subbands in the frequency domain, and a corresponding dynamic range compression gain is calculated for each virtual frequency subband.
[0028] This step is used to calculate the dynamic range compression gain, and the purpose is to amplify the sound of different intensities nonlinearly according to the hearing loss characteristics of the user, so as to realize the compensation effect of “clearly hearing small sound and not harshly hearing loud sound”. In order to realize the frequency selective compensation under the single channel architecture, the virtual multi-band processing technology is adopted in the embodiments of the present application.
[0029] Firstly, the spectrum is virtually segmented according to a preset frequency band, and a plurality of virtual frequency subbands are obtained. For example, based on the equivalent rectangular bandwidth scale which is more consistent with the frequency resolution of the human ear, the frequency range of 0-8000 Hz can be divided into four subbands: subband 1 (250-750 Hz), subband 2 (750-2000 Hz), subband 3 (2000-4000 Hz) and subband 4 (4000-8000 Hz).
[0030] Secondly, the sound pressure level can be calculated for the virtual frequency subband, and the preset compression gain rule matched with the sound pressure level is used to calculate the dynamic range compression gain corresponding to each virtual frequency subband.
[0031] Specifically, the preset compression gain rule can include: for the virtual frequency subband with a sound pressure level less than a compression starting threshold, a fixed linear gain is executed; for the virtual frequency subband with a sound pressure level reaching the compression starting threshold and less than a compression ending threshold, a smooth transition gain based on a preset compression ratio is executed; and for the virtual frequency subband with a sound pressure level reaching the compression ending threshold, a fixed ratio attenuation is executed.
[0032] For example, for sub-band 2 (750-2000 Hz), the compression start threshold can be preset as 50 dB SPL, and the compression end threshold can be preset as 100 dB SPL. When the signal energy of this sub-band is lower than 50 dB, a fixed linear gain is provided; when the signal energy of this sub-band is higher than 100 dB, a fixed proportional attenuation is provided; when the signal energy is between 50 dB and 100 dB, the gain decreases as the input energy increases, and the decreasing rate is determined by the compression ratio of 2.5:1. That is, for every 2.5 dB increase in the input signal, the output signal only increases by 1 dB. By setting different compression start, end thresholds and compression ratios (e.g., 45 / 50 / 55 / 60 dB and 2.0 / 2.5 / 3.0 / 2.8, respectively) for each virtual sub-band (sub-band 1, sub-band 2, sub-band 3, sub-band 4), precise compensation for different frequency hearing loss of the user can be achieved. In specific applications, in order to make the gain change more smooth to avoid abrupt hearing, a "soft corner" is usually set near the compression start / end threshold, so that the transition of the gain curve from the linear region to the compression region is smoother. After calculating the gain value of each sub-band, a smooth dynamic range compression gain curve covering all frequency points can be obtained by interpolation or other methods.
[0033] S102-2: Solve the background noise of the frequency spectrum by using the hybrid noise power spectral density algorithm, to analyze the adaptive noise reduction gain for suppressing the background noise.
[0034] This step is used to calculate the adaptive noise reduction gain, aiming to suppress the background noise in the environment to improve the intelligibility of the speech signal. Then, first, the characteristics of the background noise need to be accurately estimated from the noisy audio signal. Specifically, the target power spectrum of the background noise can be estimated based on the minimum statistical noise estimation algorithm and the active speech detection algorithm, which can include: based on the minimum statistical noise estimation algorithm, continuously tracking the power minimum value of each frequency point in a preset time window to obtain the initial power spectrum of the background noise; and based on the active speech detection algorithm, performing speech activity detection on the current signal frame to determine whether it contains speech components; if the current signal frame does not contain speech components, the initial power spectrum is corrected for deviation to obtain the updated target power spectrum of the background noise; if the current signal frame contains speech components, the initial power spectrum is taken as the target power spectrum of the background noise.
[0035] Among them, the basic idea of the minimum statistical noise estimation algorithm is: in a long time window (for example, several seconds), the power minimum value of each frequency point is continuously tracked. Since the speech signal usually has time-varying and high energy characteristics, and many background noises (such as fan sound and air conditioner sound) are relatively stable and low in energy, the power of the signal will decrease to the level close to the noise during the speech gap. By tracking the minimum value of the power, a robust estimation of the noise power spectrum can be obtained.
[0036] In order to further improve the accuracy of noise estimation, the speech activity detection algorithm is introduced in the embodiment of the application. In actual application, speech activity detection is performed on the current signal frame to determine whether the frame contains speech components. The speech activity detection algorithm can be based on various target features, such as the posterior signal-to-noise ratio, the spectral flatness, and the spectral entropy. It can be understood that the signal frame with high posterior signal-to-noise ratio, low spectral flatness, and low spectral entropy is more likely to contain speech. Therefore, the embodiment of the application proposes a judgment rule: according to the preset condition threshold matched with each target feature, it is judged whether each target feature meets the speech activity condition; when each target feature meets the speech activity condition, or the previous signal frame of the current signal frame contains speech components and the posterior signal-to-noise ratio meets the speech activity condition, it is considered that the current signal frame contains speech components.
[0037] For example, the preset condition threshold for the posterior signal-to-noise ratio is 2.0, the preset condition threshold for the spectral flatness is 0.3, and the preset condition threshold for the spectral entropy is 0.7. Then, when the posterior signal-to-noise ratio > 2.0, the spectral flatness < 0.3, and the spectral entropy < 0.7, it is considered that the current signal frame contains speech components. In order to avoid judgment errors, in another case, when the previous signal frame contains speech components, only the posterior signal-to-noise ratio > 1.5 is required to consider that the current signal frame contains speech components.
[0038] Further, according to the result of the speech activity detection, if the current signal frame is judged as not containing speech components (i.e. pure noise or noise dominant), the minimum statistical tracking is performed and the power spectrum estimation of the background noise is updated; otherwise, if speech is detected, the current update is skipped. This mechanism can effectively prevent the energy of the speech signal from being incorrectly counted into the noise model, thereby avoiding excessive suppression of the speech signal.
[0039] Further, after obtaining the accurate background noise power spectrum estimation, based on the target power spectrum, the optimal modified log spectrum amplitude estimation algorithm is used to calculate the adaptive noise reduction gain.
[0040] Specifically, a ratio of a power spectrum of the noisy speech to a power spectrum of the background noise is calculated to obtain a posterior SNR, and a ratio of a power spectrum of the clean speech to the power spectrum of the background noise is calculated to obtain an a priori SNR; a speech presence probability is determined according to the posterior SNR and the a priori SNR, and the optimal gain for each frequency point is dynamically adjusted based on the speech presence probability, so that the speech component obtains a high gain and the noise component obtains a low gain.
[0041] It can be understood that the optimal modified log spectrum amplitude estimation algorithm, as an advanced spectral subtraction variant, calculates a gain value between 0 and 1 for each frequency point by calculating the a priori SNR and the posterior SNR. The gain value is essentially an attenuation coefficient, and the size of the attenuation coefficient is related to the SNR of the frequency point: the frequency point with a high SNR (which is likely to be a speech component) obtains a gain close to 1 and is not substantially attenuated; and the frequency point with a low SNR (which is likely to be a noise component) obtains a gain close to 0 and is substantially suppressed.
[0042] In an optional embodiment, the adaptive noise reduction gain can also be subjected to spectral smoothing processing and / or time smoothing processing to suppress musical noise.
[0043] In order to improve the sound quality after noise reduction and reduce the "musical noise" (i.e. the residual artifacts after processing that are similar to short tones in hearing) that is easily introduced by spectral subtraction algorithms, the embodiments of the present application also perform smoothing processing on the calculated adaptive noise reduction gain, including spectral smoothing processing and time smoothing processing. The spectral smoothing processing refers to smoothing the gain values of adjacent frequency points, for example, using a small moving average filter, so that the gain curve does not appear to be jittering sharply. The time smoothing processing refers to smoothing the gain values of the same frequency point over consecutive time frames, for example, using a first-order infinite impulse response filter, so that the gain changes more smoothly. The two kinds of smoothing processing can effectively suppress musical noise and make the output sound more natural. The gain obtained after smoothing processing, i.e. the adaptive noise reduction gain, is used as the output to perform subsequent steps.
[0044] In step S103, a synthesis manner of the dynamic range compression gain and the adaptive noise reduction gain is determined according to the environmental parameter of the single-channel audio signal, and a comprehensive gain is generated based on the synthesis manner.
[0045] The environmental parameter selected by the embodiments of the present application is the average SNR of the current signal frame, which can be calculated from the previously calculated signal power spectrum and the estimated noise power spectrum. Further, the synthesis weights of the dynamic range compression gain and the adaptive noise reduction gain can be determined based on the preset synthesis rule matched with the average SNR, and the dynamic range compression gain and the adaptive noise reduction gain are weighted and summed according to the synthesis weights to obtain the comprehensive gain.
[0046] Specifically, the preset synthesis rule can comprise: assigning a higher weight to the dynamic range compression gain when the average signal-to-noise ratio is higher than a first preset threshold; assigning a higher weight to the adaptive noise reduction gain when the average signal-to-noise ratio is lower than a second preset threshold; and the first preset threshold is greater than the second preset threshold.
[0047] For example, the first preset threshold is preferably 20, and the second preset threshold is preferably 10. When the average signal-to-noise ratio is higher than 20, it indicates that the current environment is very quiet, at which time the naturalness and comfort of the sound quality should be prioritized, and unnecessary noise reduction processing should be avoided to introduce distortion, at which time the dynamic range compression gain and the adaptive noise reduction gain weight ratio can be set to [0.8:0.2]; when the average signal-to-noise ratio is higher than 10 but not higher than 20, it indicates that the environment has moderate noise, at which time a balance between sound quality and clarity needs to be achieved, and the dynamic range compression gain and the adaptive noise reduction gain weight ratio can be set to [0.5:0.5]; and when the average signal-to-noise ratio is not higher than 10, it indicates that the environment is very noisy, at which time the intelligibility of speech should be prioritized, and therefore the dynamic range compression gain and the adaptive noise reduction gain weight ratio can be set to [0.2:0.8].
[0048] After the synthesis weight is determined, the comprehensive gain can be calculated, and the calculation formula can be:
[0049] wherein, is the index of the frequency point; is the comprehensive gain of the th frequency point; and are the dynamic range compression gain and the adaptive noise reduction gain of the frequency point, respectively; and , respectively, represent the synthesis weights of the dynamic range compression gain and the adaptive noise reduction gain. It should be noted that since the dynamic range compression gain is initially calculated in subbands, in actual application, it can be expanded to have a corresponding gain value for each frequency point through interpolation or the like, so as to participate in the weighted summation here.
[0050] In step S104, the comprehensive gain is applied to the spectral representation, and the updated spectral representation is subjected to frequency-time inverse transformation, so as to reconstruct the target audio signal for output.
[0051] After obtaining the comprehensive gain, it is applied to the original spectral representation point by point, while changing the amplitude of the spectrum and retaining the phase, so as to obtain the updated spectral representation, and the calculation formula is as follows:
[0052] wherein, is the updated spectral representation; to the original spectral representation; = 0, 1, 2, …, N-1, representing the index of discrete frequency points in the frequency domain; representing the overall gain function.
[0053] Further, by performing an inverse fast Fourier transform on the gain-modified spectrum, converting it from the frequency domain back to the time domain, a processed time-domain signal frame is obtained. Since the previous frame division with 50% overlap is adopted, the continuous output signal needs to be reconstructed by overlap-add method at this time. Specifically, the latter half of the current processed signal frame is added to the former half of the next processed signal frame, which can be seamlessly spliced into a continuous output signal. This final output signal is the processed target audio signal, which not only compensates for the user's hearing loss, but also adaptively performs noise reduction according to the environmental noise, thereby providing an optimized listening experience in various environments.
[0054] In an optional embodiment, on the basis of steps S101-S104, an additional computational complexity adaptive control mechanism is added, aiming to dynamically balance between signal processing performance and device power consumption according to the real-time battery power of the single-channel hearing aid, thereby prolonging the effective endurance time of the hearing aid. Specifically, according to the obtained power information, the core parameters of the above signal processing method are periodically adjusted.
[0055] Specifically, based on the real-time battery power of the single-channel hearing aid, multiple performance modes can be preset, each mode corresponding to a different set of algorithm parameter configurations, which differ in computational complexity and processing effect: High-performance mode: when the real-time battery power is higher than a higher threshold (for example, greater than 50%), the most complete algorithm configuration is adopted to provide the best sound quality and noise reduction effect. For example, the same parameters as in the above embodiment can be used: the fast Fourier transform point number of time-frequency transform is 256 points, the virtual frequency subband is divided into 4 or more subbands, and the spectral smoothing and time smoothing processing for suppressing music noise is enabled when calculating the adaptive noise reduction gain.
[0056] Balanced mode: when the real-time battery power is in an intermediate range (for example, between 20% and 50%), the algorithm computational complexity can be appropriately reduced to save power. For example, the fast Fourier transform point number of 256 points and the number of 4 virtual frequency subbands can be maintained unchanged to ensure basic frequency resolution and compensation accuracy, but some additional functions with high computational overhead can be simplified or closed, such as reducing the filter order for time smoothing processing when calculating the adaptive noise reduction gain, or completely closing the spectral smoothing processing.
[0057] Low power mode: when the battery level is below a lower threshold (e.g. less than or equal to 20%), the remaining usage time is maximized by adopting a lower complexity core algorithm configuration. For example, the fast Fourier transform point number of the time-frequency transform can be reduced to 128 points, which will significantly reduce the amount of calculation; correspondingly, the number of virtual frequency bands can also be reduced to 3. At the same time, the music noise suppression function can be completely disabled, or even switched to a simpler noise reduction algorithm (such as the basic spectral subtraction).
[0058] In practical applications, the battery level can be queried according to a preset period, and seamless switching between different performance modes can be performed according to the above rules. It can be understood that such switching can be instantaneous or gradual to avoid sudden changes in sound quality perceived by the user. By introducing this calculation complexity adaptive control mechanism, the practicability and reliability of the single-channel hearing aid are significantly improved. When the battery level is sufficient, the user can enjoy the best hearing experience; when the battery level is about to be depleted, the core function can be ensured to continue to run by sacrificing part of the performance, avoiding the inconvenience and distress caused by the sudden shutdown of the hearing aid to the user, which has very important practical value for users who highly depend on the hearing aid.
[0059] It should be noted that in practical applications, the various threshold parameters proposed in the above embodiments can be set based on actual needs, and the present application does not limit this.
[0060] The embodiment of the present application provides a signal processing method of a single-channel hearing aid, which has at least the following beneficial effects: 1) By dynamically determining the synthesis method of the dynamic range compression gain and the adaptive noise reduction gain according to the signal-to-noise ratio and other environmental parameters, the fixed gain synthesis strategy in the prior art is solved. In a quiet environment with high signal-to-noise ratio, dynamic range compression processing can be focused on to ensure the naturalness and comfort of the sound quality; in a noisy environment with low signal-to-noise ratio, noise reduction processing can be focused on to improve the intelligibility of speech. This environment-adaptive intelligent balancing strategy can significantly optimize the overall listening experience of the user in different acoustic environments.
[0061] 2) By implementing virtual multi-band processing on a single-channel hardware architecture, frequency-selective hearing compensation close to a multi-channel hearing aid and accurate dynamic range control are realized with low hardware cost and power consumption, and combined with an adaptive noise reduction algorithm, environmental noise can be effectively suppressed while reducing the distortion caused by noise reduction processing, so that the processed sound is more natural, pleasant and comfortable to wear.
[0062] Further, as a specific implementation of Figure 1 , the embodiment of the present application provides a signal processing device of a single-channel hearing aid, such as Figure 2As shown, the device can include: a signal acquisition unit 210, a data analysis unit 220, a gain synthesis unit 230, and a reconstruction output unit 240.
[0063] The signal acquisition unit 210 can be used to acquire a time domain signal frame of a single-channel audio signal, and perform time-frequency transformation on the time domain signal frame to obtain a spectral representation of the time domain signal frame in the frequency domain; The data analysis unit 220 can be used to divide the spectrum into a plurality of virtual frequency subbands in the frequency domain, calculate a corresponding dynamic range compression gain for each virtual frequency subband, and solve the background noise of the spectrum using a hybrid noise power spectral density algorithm to analyze an adaptive noise reduction gain for suppressing the background noise; The gain synthesis unit 230 can be used to determine a synthesis manner of the dynamic range compression gain and the adaptive noise reduction gain according to an environmental parameter of the single-channel audio signal, and generate a comprehensive gain based on the synthesis manner; The reconstruction output unit 240 can be used to apply the comprehensive gain to the spectral representation, and perform frequency-time inverse transformation on the updated spectral representation to reconstruct a target audio signal for output.
[0064] It should be noted that other corresponding descriptions of the functions of the signal processing device of the single-channel hearing aid provided in the embodiments of the present application can be referred to Figure 1 The corresponding description of the method is not repeated here.
[0065] Further, the embodiments of the present application also provide a single-channel hearing aid, as Figure 3 As shown, the single-channel hearing aid can at least include a sound acquisition module 310, an analog-to-digital conversion module 320, a core processing module 330, a digital-to-analog conversion module 340, and a sound output module 350 connected in sequence.
[0066] The sound output module 310 can be used to capture environmental sound information and convert it into an initial electrical signal. Preferably, an omnidirectional MEMS microphone can be used, with a frequency response range of 100Hz-10kHz and a signal-to-noise ratio of >65dB.
[0067] The analog-to-digital conversion module 320 can be used to perform analog-to-digital conversion processing on the electrical signal to obtain an original single-channel audio signal. Preferably, a low-power, high-precision 24-bit sigma-delta ADC can be used, with a sampling rate of 16-48kHz adjustable and a dynamic range of >100dB.
[0068] The core processing module 330 includes a signal processing device of a single-channel hearing aid as Figure 2 shown, which can be used to execute a signal processing method of a single-channel hearing aid as Figure 1 shown.
[0069] The digital-to-analog conversion module 340 can be configured to perform digital-to-analog conversion on the target audio signal output by the core processing module to obtain a target electrical signal. Preferably, a high-precision sigma-delta DAC can be used, with total harmonic distortion < 0.01%.
[0070] The sound output module 350 can be configured to convert the target electrical signal into a sound wave for output. Preferably, a sensitive microphone can be used, with a frequency response of 200 Hz-8 kHz.
[0071] In an optional embodiment, the single-channel hearing aid can further include a power management module (not shown in the figure), which can be configured to supply power to the entire hearing aid system. Preferably, a rechargeable micro lithium battery can be used, supporting a low-power sleep mode.
[0072] It should be noted that other corresponding descriptions of the functions of the various functional modules of the single-channel hearing aid provided in the embodiments of the present application can be referred to the corresponding descriptions of the methods shown in the embodiments of the present application, which will not be repeated here. Figure 1
[0073] Those skilled in the art can clearly understand the specific working process of the system, device, module and unit described above, which can refer to the corresponding process in the foregoing method embodiments, and for the sake of brevity, will not be repeated here.
[0074] In addition, the functional units in each of the embodiments of the present application can be physically independent of each other, or two or more functional units can be integrated together, and all functional units can be integrated in one processing unit. The integrated functional units can be realized in the form of hardware or in the form of software or firmware.
[0075] Those skilled in the art can understand that the integrated functional units, if realized in the form of software and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product in essence or all or part of the technical solutions. The computer software product is stored in a storage medium, which includes a plurality of instructions for causing a computing device (such as a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the embodiments of the present application when the instructions are run. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various storage medium that can store program codes.
[0076] Alternatively, all or part of the steps of the foregoing method embodiments can be accomplished by program instruction related hardware (such as a computing device of a personal computer, a server, or a network device, etc.), which can be stored in a computer readable storage medium, and when the program instruction is executed by a processor of the computing device, the computing device executes all or part of the steps of the method according to the embodiments of the present application.
[0077] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand: within the spirit and principles of the present application, the technical solutions recorded in the foregoing embodiments can still be modified, or part or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the corresponding technical solutions deviate from the protection scope of the present application.
Claims
1. A signal processing method for a single-channel hearing aid, characterized by, The method comprises: obtaining a time-domain signal frame of a single-channel audio signal, and performing time-frequency conversion on the time-domain signal frame to obtain a spectral representation of the time-domain signal frame in a frequency domain; in the frequency domain, dividing the spectrum into a plurality of virtual frequency subbands, calculating a corresponding dynamic range compression gain for each virtual frequency subband, and solving background noise of the spectrum by using a hybrid noise power spectral density algorithm to analyze an adaptive noise reduction gain for suppressing the background noise; determining a combination mode of the dynamic range compression gain and the adaptive noise reduction gain according to an environmental parameter of the single-channel audio signal, generating a comprehensive gain based on the combination mode; applying the comprehensive gain to the spectral representation, and performing frequency-time inverse conversion on the updated spectral representation to reconstruct a target audio signal for output.
2. The method of claim 1, wherein, The method comprises: obtaining a time-domain signal frame of a single-channel audio signal, and performing time-frequency conversion on the time-domain signal frame to obtain a spectral representation of the time-domain signal frame in a frequency domain; in the frequency domain, dividing the spectrum into a plurality of virtual frequency subbands, calculating a corresponding dynamic range compression gain for each virtual frequency subband, and solving background noise of the spectrum by using a hybrid noise power spectral density algorithm to analyze an adaptive noise reduction gain for suppressing the background noise; 3. The method of claim 1, wherein, determining a combination mode of the dynamic range compression gain and the adaptive noise reduction gain according to an environmental parameter of the single-channel audio signal, generating a comprehensive gain based on the combination mode; applying the comprehensive gain to the spectral representation, and performing frequency-time inverse conversion on the updated spectral representation to reconstruct a target audio signal for output. The method comprises: obtaining a time-domain signal frame of a single-channel audio signal, and performing time-frequency conversion on the time-domain signal frame to obtain a spectral representation of the time-domain signal frame in a frequency domain; in the frequency domain, dividing the spectrum into a plurality of virtual frequency subbands, calculating a corresponding dynamic range compression gain for each virtual frequency subband, and solving background noise of the spectrum by using a hybrid noise power spectral density algorithm to analyze an adaptive noise reduction gain for suppressing the background noise; determining a combination mode of the dynamic range compression gain and the adaptive noise reduction gain according to an environmental parameter of the single-channel audio signal, generating a comprehensive gain based on the combination mode; applying the comprehensive gain to the spectral representation, and performing frequency-time inverse conversion on the updated spectral representation to reconstruct a target audio signal for output.
4. The method of claim 1, wherein, The method comprises: obtaining a time-domain signal frame of a single-channel audio signal, and performing time-frequency conversion on the time-domain signal frame to obtain a spectral representation of the time-domain signal frame in a frequency domain; in the frequency domain, dividing the spectrum into a plurality of virtual frequency subbands, calculating a corresponding dynamic range compression gain for each virtual frequency subband, and solving background noise of the spectrum by using a hybrid noise power spectral density algorithm to analyze an adaptive noise reduction gain for suppressing the background noise; 5. The method of claim 4, wherein, determining a combination mode of the dynamic range compression gain and the adaptive noise reduction gain according to an environmental parameter of the single-channel audio signal, generating a comprehensive gain based on the combination mode; applying the comprehensive gain to the spectral representation, and performing frequency-time inverse conversion on the updated spectral representation to reconstruct a target audio signal for output. The method comprises: obtaining a time-domain signal frame of a single-channel audio signal, and performing time-frequency conversion on the time-domain signal frame to obtain a spectral representation of the time-domain signal frame in a frequency domain; in the frequency domain, dividing the spectrum into a plurality of virtual frequency subbands, calculating a corresponding dynamic range compression gain for each virtual frequency subband, and solving background noise of the spectrum by using a hybrid noise power spectral density algorithm to analyze an adaptive noise reduction gain for suppressing the background noise; determining a combination mode of the dynamic range compression gain and the adaptive noise reduction gain according to an environmental parameter of the single-channel audio signal, generating a comprehensive gain based on the combination mode; applying the comprehensive gain to the spectral representation, and performing frequency-time inverse conversion on the updated spectral representation to reconstruct a target audio signal for output. The method comprises: obtaining a time-domain signal frame of a single-channel audio signal, and performing time-frequency conversion on the time-domain signal frame to obtain a spectral representation of the time-domain signal frame in a frequency domain; in the frequency domain, dividing the spectrum into a plurality of virtual frequency subbands, calculating a corresponding dynamic range compression gain for each virtual frequency subband, and solving background noise of the spectrum by using a hybrid noise power spectral density algorithm to analyze an adaptive noise reduction gain for suppressing the background noise; determining a combination mode of the dynamic range compression gain and the adaptive noise reduction gain according to an environmental parameter of the single-channel audio signal, generating a comprehensive gain based on the combination mode; applying the comprehensive gain to the spectral representation, and performing frequency-time inverse conversion on the updated spectral representation to reconstruct a target audio signal for output. The method comprises: obtaining a time-domain signal frame of a single-channel audio signal, and performing time-frequency conversion on the time-domain signal frame to obtain a spectral representation of the time-domain signal frame in a frequency domain; in the frequency domain, dividing the spectrum into a plurality of virtual frequency subbands, calculating a corresponding dynamic range compression gain for each virtual frequency subband, and solving background noise of the spectrum by using a hybrid noise power spectral density algorithm to analyze an adaptive noise reduction gain for suppressing the background noise; determining a combination mode of the dynamic range compression gain and the adaptive noise reduction gain according to an environmental parameter of the single-channel audio signal, generating a comprehensive gain based on the combination mode; applying the comprehensive gain to the spectral representation, and performing frequency-time inverse conversion on the updated spectral representation to reconstruct a target audio signal for output.
6. The method of claim 5, wherein, The voice activity detection algorithm is used to detect voice activity of the current signal frame to determine whether the current signal frame contains voice components, and the method comprises the following steps: An active voice detection algorithm based on multiple target features is used to calculate the posterior signal-to-noise ratio, spectral flatness and spectral entropy of the current signal frame; A preset condition threshold is determined for each target feature to determine whether the target feature meets the voice activity condition; When each target feature meets the voice activity condition, or the previous signal frame of the current signal frame contains voice components and the posterior signal-to-noise ratio meets the voice activity condition, it is determined that the current signal frame contains voice components.
7. The method of claim 4, wherein, The optimal modified log spectrum amplitude estimation algorithm is used to calculate the adaptive noise reduction gain based on the power spectrum of the background noise, and the method comprises the following steps: The ratio of the noisy speech power spectrum to the power spectrum of the background noise is calculated to obtain the posterior signal-to-noise ratio, and the ratio of the clean speech power spectrum to the power spectrum of the background noise is calculated to obtain the prior signal-to-noise ratio; The voice presence probability is determined based on the posterior signal-to-noise ratio and the prior signal-to-noise ratio, and the optimal gain for each frequency point is dynamically adjusted based on the voice presence probability, so that the voice components obtain a high gain and the noise components obtain a low gain.
8. The method of claim 1, wherein, The synthesis mode of the dynamic range compression gain and the adaptive noise reduction gain is determined based on the environmental parameters of the single-channel audio signal, and the comprehensive gain is generated based on the synthesis mode, and the method comprises the following steps: The average signal-to-noise ratio of the single-channel audio signal is calculated, the synthesis weights of the dynamic range compression gain and the adaptive noise reduction gain are determined based on a preset synthesis rule matched with the average signal-to-noise ratio, and the dynamic range compression gain and the adaptive noise reduction gain are weighted and summed according to the synthesis weights to obtain the comprehensive gain; The preset synthesis rule comprises: when the average signal-to-noise ratio is higher than a first preset threshold, a higher weight is assigned to the dynamic range compression gain; when the average signal-to-noise ratio is lower than a second preset threshold, a higher weight is assigned to the adaptive noise reduction gain; and the first preset threshold is greater than the second preset threshold.
9. The method according to any one of claims 1 to 8, characterized in that, The method further comprises: The adaptive noise reduction gain is subjected to spectral smoothing processing and / or time smoothing processing to suppress musical noise; The spectral smoothing processing comprises smoothing the gain values of adjacent frequency points, and the time smoothing processing comprises smoothing the gain values of the same frequency point in consecutive time frames.
10. A single channel hearing aid, characterized by The single-channel hearing aid at least comprises sound acquisition module, analog-to-digital conversion module, core processing module, digital-to-analog conversion module and sound output module connected in sequence; The sound output module is used to capture environmental sound information and convert it into an initial electrical signal; The analog-to-digital conversion module is used to perform analog-to-digital conversion processing on the electrical signal to obtain an original single-channel audio signal; The core processing module is used to execute the signal processing method of the single-channel hearing aid according to any one of claims 1-9; The digital-to-analog conversion module is used to perform digital-to-analog conversion processing on the target audio signal output by the core processing module to obtain a target electrical signal; The sound output module is configured to convert the target electrical signal into a sound wave for output.