Binaural neural synchronous hearing aid system

By optimizing Bluetooth transmission, deep learning, and adaptive filtering algorithms, combined with high-energy-density batteries and nano-coating technology, the problems of signal synchronization, environmental adaptability, and device durability in binaural neural synchronous hearing aid systems have been solved, achieving high-precision synchronization, high-quality sound, and widely adaptable hearing rehabilitation effects.

CN123123799APending Publication Date: 2026-11-17ZUODIAN IND (HUBEI) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202510622756.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-05-15
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing binaural neural synchronous hearing aid systems have shortcomings in signal synchronization, adaptability to complex acoustic environments, personalized adjustment, and device durability, which affect the auditory experience and applicability of people with hearing impairments.

Method used

Employing a high-precision synchronization module, an intelligent acoustic processing module, a low-latency high-fidelity processing module, an intelligent personalized debugging module, and a durable battery life module, combined with deep learning, adaptive filtering algorithms, parallel processing architecture, nano-coating technology, and high-energy-density batteries, it optimizes Bluetooth transmission, signal processing, and device design, providing personalized debugging and a wide range of compatibility solutions.

Benefits of technology

It achieves high-precision signal synchronization, strong environmental adaptability, excellent auditory experience, convenient debugging, and long-lasting battery life, expanding the range of applicable people and improving the effect of hearing rehabilitation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123799A_ABST
    Figure CN123123799A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of hearing aid technology, specifically a binaural neural synchronous hearing aid system. It includes a high-precision synchronization module that achieves microsecond-level synchronization based on optimized Bluetooth protocol and Kalman filter calibration; an intelligent acoustic processing module that integrates deep learning and adaptive filtering to improve sound processing capabilities in complex environments; a low-latency, high-fidelity processing module that ensures sound quality and real-time performance through parallel architecture and advanced coding; an intelligent personalized adjustment module that utilizes an app and machine learning for rapid and accurate adjustment; a durable battery module that enhances durability with a nano-coating and high-energy battery; and a collaborative adaptation module that provides fitting solutions for individuals with different levels of hearing loss. Verified in multiple scenarios, this invention significantly improves hearing aid performance and user experience, effectively expanding the range of applicable users.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of hearing aid technology, specifically a binaural neural synchronous hearing aid system. Background Technology

[0002] With the accelerating global aging process and the impact of factors such as noise pollution, the number of people with hearing impairments continues to grow. Binaural neural synchronous hearing aid systems, as important devices to help hearing-impaired individuals restore their auditory function, are designed to simulate the natural binaural effect, achieving accurate sound source localization and clear sound perception. However, existing technologies face many problems that urgently need to be addressed.

[0003] In terms of binaural signal synchronization, the traditional Bluetooth transmission mechanism is a key factor leading to insufficient synchronization accuracy. During Bluetooth transmission, the signal undergoes multiple stages, including encoding, modulation, wireless transmission, demodulation, and decoding. Each stage introduces latency, with cumulative latency reaching tens of milliseconds. Simultaneously, the difference in algorithm processing speed between the left and right hearing aids further exacerbates the sound asynchrony. For example, in dynamic sound source scenarios, the arrival time of a moving speaker's voice in the left and right ears should have a slight difference. However, the device latency and processing speed differences interfere with this natural time difference perception, causing confusion in the user's sense of direction and making it impossible to accurately determine the sound source's location.

[0004] Adaptability to complex acoustic environments is also a significant challenge. In multi-source environments such as conference rooms and restaurants, as well as in environments with strong background noise like traffic roads and shopping malls, existing noise suppression and speech enhancement algorithms struggle to effectively separate target sounds. From a technical perspective, traditional algorithms have limited processing capabilities for low-frequency sounds and cannot effectively suppress low-frequency background noise. Furthermore, their dynamic range compression strategies are often flawed, causing faint target sounds (such as whispers within 3 meters) to be masked by background noise. In addition, some algorithms are prone to misjudging complex, overlapping audio signals, treating some target sounds as noise and reducing speech intelligibility.

[0005] Signal processing latency, like audio quality loss, negatively impacts user experience. To achieve binaural processing, the system requires extensive signal processing operations, such as encoding / decoding conversion and noise reduction algorithms. These operations introduce additional latency, causing users to experience rhythm perception distortion when listening to music or engaging in conversations. Regarding audio quality, common Bluetooth transmission protocols (such as SBC and AAC) compress audio signals to meet bandwidth requirements when transmitting high-resolution audio, resulting in loss of sound detail and reduced realism and clarity.

[0006] Regarding personalized adjustments, the system requires precise calibration due to the significant differences in hearing curves and lifestyle needs among different users. However, in reality, the skill levels of some audiologists vary, and they lack professional adjustment tools (such as real-ear analysis equipment), leading to unreasonable gain distribution, resulting in problems such as excessive amplification of high frequencies causing a harsh sound and imbalance in volume between the two ears.

[0007] Furthermore, the maintenance and battery life of these devices are significant challenges. The delicate internal circuitry of hearing aids is constantly exposed to the user's environment, making it susceptible to corrosion from sweat and dust, potentially leading to circuit malfunctions and requiring regular factory maintenance. To maintain binaural synchronous operation and Bluetooth communication, battery power consumption is high; some models have a battery life of less than 8 hours when Bluetooth is enabled, causing users the inconvenience of frequent battery replacements or carrying a charging case. For patients with profound hearing loss or central auditory processing disorders, the existing systems are not well-suited; the reliability of the devices is also difficult to guarantee in extreme physical environments; users with asymmetrical hearing loss also face fitting difficulties. These problems severely restrict the application and development of binaural neural synchronous hearing aid systems. Summary of the Invention

[0008] To address the problems in the prior art, this invention provides a binaural neural synchronous hearing aid system, specifically a binaural neural synchronous hearing aid system that can solve multiple problems such as binaural signal synchronization, adaptability to complex environments, and personalized adjustment, thereby improving the auditory experience of people with hearing impairments through the integration of multiple technologies.

[0009] The technical solution adopted by this invention to solve its technical problem is: a binaural neural synchronous hearing aid system, comprising: A high-precision synchronization module is used to optimize the Bluetooth transmission protocol and employ a dynamic time calibration algorithm to achieve high-precision synchronization of binaural signals. The intelligent acoustic processing module combines deep learning and adaptive filtering algorithms for noise suppression and speech enhancement. The low-latency, high-fidelity processing module reduces signal processing latency and ensures sound quality through a parallel processing architecture and efficient encoding / decoding technology. The intelligent personalized debugging module utilizes mobile terminal APP and machine learning algorithms to achieve personalized debugging. The durable battery module uses nano-coating waterproof and dustproof technology and high-energy-density batteries to improve the device's durability and battery life. The collaborative adaptation module provides adaptation solutions for users with different types of hearing loss.

[0010] Specifically, in the high-precision synchronization module, the Bluetooth transmission protocol optimizes the data packet structure, employs time-division multiplexing technology, and the dynamic time calibration algorithm is based on the Kalman filter principle, using the formula... , Where X(k) is the state vector, A is the state transition matrix, B is the control matrix, U(k) is the control vector, W(k) is the process noise, Z(k) is the measurement vector, H is the measurement matrix, and V(k) is the measurement noise. The transmission time of the left and right ear signals is monitored and adjusted in real time to keep the synchronization error within 1 millisecond.

[0011] Specifically, the intelligent acoustic processing module is built based on a convolutional neural network and an adaptive filtering algorithm. The convolutional neural network contains no fewer than 5 convolutional layers, 3 pooling layers, and 2 fully connected layers. It identifies sound features through training on a large amount of acoustic data. The adaptive filtering algorithm uses a low-frequency boosting filtering formula. It processes low-frequency sounds and improves the accuracy of target sound recognition to over 90% in complex acoustic environments.

[0012] Specifically, the low-latency high-fidelity processing module adopts a parallel processing architecture, with the left and right ear signal processing units sampling synchronously at a sampling frequency of no less than 48kHz and a quantization bit depth of no less than 24 bits. The encoding and decoding unit adopts lossless audio coding technology based on wavelet transform, with an encoding and decoding delay of no more than 2 milliseconds. The noise reduction algorithm adopts an optimized Wiener filtering algorithm.

[0013] Specifically, the mobile terminal APP of the intelligent personalized tuning module allows users to input hearing curve data and select life scene modes. It generates personalized solutions using a tuning model based on support vector regression algorithm, while also supporting manual fine-tuning of parameters and automatically optimizing the solution based on user auditory feedback, reducing the tuning time to less than 5 minutes.

[0014] Specifically, in the durable battery life module, the nano-coating waterproof and dustproof technology forms a superhydrophobic and superoleophobic coating with a thickness of 30-80 nanometers on the shell surface, which, combined with the sealed structure design, achieves an IP67 protection level; it uses a lithium-sulfur battery with an energy density of not less than 350Wh / kg, and through an intelligent power management algorithm, it monitors the operating status of each module in real time. When the Bluetooth connection is idle for more than a preset time (1-2 minutes), it automatically switches to a low-power sleep mode, reducing the power consumption of the Bluetooth module by 70%-90%, so that the battery life of the device can be extended to more than 20 hours when the Bluetooth function is turned on.

[0015] Specifically, the collaborative adaptation module is designed for patients with profound hearing loss. It features a dedicated cochlear implant collaborative working interface and a unified synchronization protocol to control the signal transmission delay between the hearing aid and the cochlear implant to within 5 milliseconds, enabling bidirectional information interaction. For patients with central auditory processing disorders, the mobile terminal APP provides at least two types of auditory training games, including sound localization and speech recognition. After three months of training, patients' speech comprehension ability can be improved by an average of 25%-35%. For users with asymmetric hearing loss, a personalized binaural balance algorithm model based on the difference in binaural hearing curves is established, which can dynamically adjust the difference in sound amplification between the two sides to between 15-25 dB and the low-frequency compensation amount to between 2-5 dB.

[0016] Specifically, during the system initialization phase, the high-precision synchronization module obtains the initial delay time of the left and right ear signal transmission by sending a test signal, which is used as the initial value of the state vector of the dynamic time calibration algorithm. During operation, the actual arrival time of the left and right ear signals is collected at a fixed time interval of 5-15 milliseconds for algorithm iterative calculation.

[0017] Specifically, the intelligent acoustic processing module preprocesses the acquired sound signals, including normalization to limit the signal amplitude to between -1 and 1, and frame processing, with each frame lasting 10-30 milliseconds. In the noise reduction process, the adaptive filter coefficients are automatically adjusted based on the noise intensity and spectral characteristics determined by the CNN model, with an adjustment step size of 0.01-0.1.

[0018] The audio signal preprocessing and CNN feature extraction include: Input stage: The analog signal collected by the microphone is converted into a digital signal by A / D conversion, and then normalized (amplitude limited to [-1,1]) and framed (each frame is 10-30 milliseconds, such as 20ms) to form a frame sequence x(n).

[0019] CNN model inference: Each frame of signal is input into a pre-trained CNN (containing ≥5 convolutional layers + 3 pooling layers), and outputs two types of key information: Noise intensity level: classified by the fully connected layer Softmax. (0 = no noise, 3 = strong noise), or directly output the noise energy percentage. .

[0020] Spectral characteristic vector: Extracts the frequency domain distribution of noise through convolutional layer feature maps, such as the energy proportion of low frequency (20-500Hz), mid frequency (500Hz-3kHz), and high frequency (>3kHz). Or the power spectral density (PSD) distribution of a specific frequency band.

[0021] Adaptive filter parameter mapping rules Let the adaptive filter be an M-order finite impulse response (FIR) filter with coefficients of...

[0022] The mapping relationship between its adjustment strategy and the CNN output is as follows: (1) Adjust the filter step size and update frequency according to the noise intensity. Step size parameters : when (Low noise): (Small step size to ensure steady-state accuracy).

[0023] when (Medium noise): (Balancing convergence speed and stability).

[0024] when (Strong noise): (Large step size, quickly tracking noise changes).

[0025] Update frequency: In noisy scenarios, the filter coefficient update interval is shortened from the default 1 frame (20ms) to 0.5 frames (10ms), improving real-time performance.

[0026] (2) Adjust the frequency response of the filter according to the spectral characteristics. Low-frequency noise dominates ( Activate the low-frequency notch filter and set the attenuation coefficient in the 20-500Hz frequency band. It suppresses low-frequency noise from air conditioners, engines, etc.

[0027] Adjust the adaptive filtering formula as follows (d is the delay order, corresponding to the period of the low-frequency noise).

[0028] High-frequency noise dominates ( Enhanced high-frequency cutoff filter, The gain of the above frequency bands is set to It suppresses high-frequency noises such as keyboard clicks and glass friction.

[0029] Broadband noise (uniform noise distribution across the entire frequency band): Enable Wiener filter enhancement mode: Estimate the signal power spectrum in real time based on the noise PSD, and adjust the filter coefficients accordingly. ,in The autocorrelation matrix of the input signal, This is the cross-correlation vector between the input and the desired signal.

[0030] Coefficient update algorithm (taking LMS as an example) Expected signal d(k): The clean speech signal predicted by the CNN model (the denoised clean speech is used as the label during training, and is approximated as the noisy signal minus the noise estimate during inference).

[0031] Error calculation: .

[0032] Coefficient update: .

[0033] in, The noise intensity is dynamically adjusted (as per the rules mentioned above). The input signal vector for the current frame.

[0034] Computational resource allocation: In the parallel processing architecture, CNN inference and filter coefficient updates are handled by independent DSP cores to avoid resource contention and ensure that the processing time for a single frame is less than 15ms (meeting the 30ms frame length requirement).

[0035] Smooth transition of coefficients: Introducing first-order inertial filtering This prevents audio stuttering caused by sudden changes in coefficients.

[0036] Specifically, the output unit of the low-latency high-fidelity processing module has a D / A conversion accuracy of no less than 24 bits, the power amplifier adopts a Class-D digital power amplifier with an efficiency of no less than 80%, and the total delay in the entire signal processing link does not exceed 5 milliseconds, and the subjective evaluation score of sound quality is no less than 8 points (out of 10).

[0037] The beneficial effects of this invention are: High-precision synchronization: Through improved Bluetooth transmission protocol and dynamic time calibration algorithm, the synchronization error of left and right ear sound is controlled within 1 millisecond, which significantly improves the accuracy of sound source localization and effectively solves the problem of user confusion in dynamic sound source scenarios.

[0038] Strong environmental adaptability: The intelligent acoustic processing module, which combines deep learning and adaptive filtering algorithms, significantly improves the accuracy of target sound recognition in complex acoustic environments. Users can clearly hear normal conversations within 3 meters and whispers within 2 meters in environments such as multi-person meetings and restaurants.

[0039] Superior listening experience: The low-latency, high-fidelity processing module adopts a parallel processing architecture and advanced encoding and decoding technology to reduce signal processing latency to less than 2 milliseconds while ensuring high-fidelity sound quality. Users experience no noticeable delay when listening to music or having conversations, and the sound details are rich.

[0040] Convenient and personalized debugging: The intelligent personalized debugging module enables fast and accurate personalized debugging through mobile terminal APP and machine learning algorithms. Users can complete the debugging process independently, which greatly shortens the debugging time and significantly improves user satisfaction after debugging.

[0041] Long and durable battery life: The nano-coating waterproof and dustproof technology and high-energy-density battery of the durable battery module enable the device to achieve IP67 protection performance and extend the battery life from 8 hours to 24 hours, reducing the user's maintenance costs and inconvenience.

[0042] Wide compatibility: The collaborative adaptation module provides effective solutions for patients with profound hearing loss, patients with central auditory processing disorders, and users with asymmetric hearing loss, expanding the range of people who can use hearing aids and improving the auditory rehabilitation effect for patients with different types of hearing impairment. Attached Figure Description

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0044] Figure 1 A block diagram of the binaural neural synchronous hearing aid system provided by the present invention. Detailed Implementation

[0045] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0046] like Figure 1 As shown, the binaural neural synchronous hearing aid system of the present invention includes the following modules: High-precision synchronization module: Employs an improved Bluetooth Low Energy transmission protocol and dynamic time calibration algorithm. Regarding the transmission protocol, the data packet structure is optimized to reduce redundant information and improve transmission efficiency. Algorithm-wise, a time calibration model is established based on the Kalman filter principle, with the following formula:

[0047]

[0048] Where X(k) is the state vector, A is the state transition matrix, B is the control matrix, U(k) is the control vector, W(k) is the process noise, Z(k) is the measurement vector, H is the measurement matrix, and V(k) is the measurement noise. High-precision synchronization is achieved by real-time monitoring and adjustment of the left and right ear signal transmission time.

[0049] Intelligent acoustic processing module: Combining deep learning's convolutional neural network (CNN) with adaptive filtering algorithms, a noise suppression and speech enhancement model is constructed. First, a large amount of acoustic data from different scenarios is collected for model training, enabling the model to recognize various sound features. In actual operation, noise reduction and enhancement parameters are automatically adjusted based on the ambient noise intensity and target sound characteristics. For low-frequency sound processing, a low-frequency boosting filtering algorithm is used, with the following formula: ,in This is the low-frequency boost factor. Adjusting this factor enhances the target low-frequency sound and suppresses low-frequency noise.

[0050] Low-latency, high-fidelity processing module: This module features a parallel processing architecture for simultaneous sampling, processing, and output of signals from both ears. The signal processing flow employs efficient encoding / decoding algorithms, such as lossless audio coding based on wavelet transform, to reduce encoding / decoding latency and sound quality loss. Simultaneously, the noise reduction algorithm's computational flow is optimized to shorten processing time.

[0051] Intelligent Personalized Tuning Module: A mobile-based intelligent tuning app has been developed, allowing users to input their hearing curve data and lifestyle needs. The system utilizes machine learning algorithms to build a tuning model based on extensive user data, automatically generating personalized gain allocation and volume balancing schemes. Simultaneously, the app allows users to manually fine-tune parameters and uses built-in sensors to monitor the usage environment in real time, automatically optimizing the tuning scheme.

[0052] Durable Battery Life Module: Employing nano-coating for waterproofing and dustproofing, and a sealed structural design, this module protects the internal precision circuitry. A new high-energy-density battery has been developed, combined with intelligent power management algorithms to dynamically adjust power consumption based on device operating status. For example, when Bluetooth connectivity is idle, the operating frequency of the Bluetooth module is reduced; when the amount of signal processing required by both ears is low, computational resource usage is minimized.

[0053] Collaborative Adaptation Module: For patients with profound hearing loss, a collaborative working interface with the cochlear implant is designed. Through synchronization protocol optimization, low-latency matching between the hearing aid and the cochlear implant is achieved. For patients with central auditory processing disorders, auxiliary training functions are developed, providing auditory training games through an app to help patients improve their auditory processing abilities. For users with asymmetrical hearing loss, a personalized binaural balance algorithm model is established, dynamically adjusting the amplification and frequency response of sound from both ears based on the differences in their hearing curves.

[0054] Example 1: Implementation of a High-Precision Synchronization Module In this embodiment, the construction of the high-precision synchronization module revolves around the optimization of the Bluetooth transmission protocol and the dynamic time calibration algorithm, aiming to completely solve the problem of insufficient synchronization accuracy of binaural signals.

[0055] First, the Bluetooth Low Energy (BLE) transmission unit was modified. Traditional Bluetooth data packets contain a large amount of redundant information. This redundancy not only occupies transmission bandwidth but also increases signal processing time, leading to latency. This embodiment redesigns the data packet structure, removing unnecessary check fields and duplicate device identification information, increasing the effective payload ratio of the data packets from 60% to 90%. Simultaneously, Time Division Multiplexing (TDM) technology is employed, dividing the Bluetooth transmission timeline into 5-millisecond time slots. The left and right hearing aids transmit data in different time slots; for example, the left ear signal is transmitted in odd-numbered time slots, and the right ear signal is transmitted in even-numbered time slots. This avoids signal collisions and mutual interference, reducing the likelihood of latency from a transmission mechanism perspective.

[0056] The dynamic time calibration unit is the core component for achieving high-precision synchronization. It is based on the Kalman filter algorithm, which optimizes the system state by establishing system state and measurement equations. During system initialization, both hearing aids simultaneously send specific test signals containing precise timestamps to the receiving end. The receiving end records the actual arrival time of the signal and compares it with the transmission timestamp to calculate the initial delay time. Treat it as a state vector The initial value.

[0057] In actual operation, the system continuously collects the actual arrival times of the left and right ear signals at 10-millisecond intervals, using this as the measurement vector Z(k). Based on the Kalman filter formula, a time update is first performed to predict the state vector for the next moment. The state transition matrix A is preset based on the system's stability and signal transmission characteristics, while the control matrix B and control vector U(k) are used to address the impact of potential external interference or human intervention on signal transmission. Next, the Kalman gain is calculated. The covariance matrix P(k|k-1) reflects the uncertainty of the predicted state, the measurement matrix H maps the state vector to the measurement space, and R is the covariance matrix of the measurement noise. The predicted state is corrected using Kalman gain to obtain the updated state vector. Simultaneously update the covariance matrix. .

[0058] Through continuous iteration of the above process, the system can monitor and adjust the signal transmission time of the left and right ears in real time. Extensive experimental verification has shown that, in various complex environments, including scenarios with multipath interference and signal obstruction, this module can stably control the synchronization error between the left and right ears to within 1 millisecond. In extreme tests with dynamic sound source movement speeds reaching 2 meters per second, the user's accuracy in judging the direction of the sound source reached 95%, an improvement of nearly 40% compared to traditional systems, effectively solving the problem of directional confusion caused by signal asynchrony.

[0059] Example 2: Application of the intelligent acoustic processing module This embodiment focuses on the application of the intelligent acoustic processing module in complex acoustic environments. Through the deep integration of deep learning and adaptive filtering algorithms, the system's ability to recognize and process target sounds is improved.

[0060] First, a voice recognition model based on a convolutional neural network (CNN) was constructed. During the data collection phase, a total of 25,000 hours of acoustic data were collected in various typical and complex scenarios, such as large conference rooms accommodating more than 50 people, shopping malls with daily foot traffic exceeding 10,000, and main traffic arteries. This data covered speech voices of different genders, ages, and accents, as well as various background noises, including air conditioner operation sounds, vehicle engine noises, and crowd noise. The collected data underwent preprocessing, including noise reduction, normalization, and annotation, clearly distinguishing between target speech and noise to form a training dataset.

[0061] The CNN model architecture comprises five convolutional layers, three pooling layers, and two fully connected layers. The convolutional layers use kernels of varying sizes, such as 3×3 and 5×5, to extract local features of the sound signal in both the time and frequency domains. For example, the first convolutional layer captures high-frequency details of the sound using a small kernel, while subsequent convolutional layers gradually increase the kernel size to obtain more macroscopic structural features of the sound. The pooling layers employ max pooling to reduce the dimensionality of the feature maps output by the convolutional layers, reducing computation while preserving key features. The fully connected layers integrate the feature vectors processed by the multiple layers and output the probability that the sound belongs to the target speech or noise using the Softmax function. After two months of training, the model achieved a 92% accuracy rate in recognizing target speech and noise on the validation set.

[0062] In actual operation, the analog sound signal captured by the microphone is first converted into a digital signal by an A / D converter, then normalized to limit the signal amplitude to the range of [-1, 1], and then divided into frames of 20 milliseconds each. Each frame of data is input into the CNN model, which determines the probability of it being target speech or noise. For signals determined to be noise, an adaptive filtering algorithm is used for processing.

[0063] When strong low-frequency noise (such as low-frequency vibration sound from an air conditioner) is detected, the low-frequency boosting filter formula is applied. Automatically adjust low-frequency boost factor Experiments have shown that when the low-frequency noise energy exceeds a certain threshold, it will... By gradually adjusting the value from the initial value of 0.2 to 0.5, low-frequency noise can be effectively suppressed while enhancing the target speech signal in the low-frequency band.

[0064] When adjusting the coefficients in low-frequency noise scenarios: Input conditions: Low-frequency vibration noise of air conditioner (center frequency 100Hz, energy percentage) CNN output, noise level L=0.3, spectral characteristics [0.8,0.1,0.1].

[0065] Adjustment process: 1. Step length The update interval is 10ms.

[0066] 2. The low-frequency boost factor starts from the initial value. Gradually increase to (through formula) ,in ).

[0067] 3. The filter attenuation at 100Hz is increased from 10dB to 25dB, while the target speech band in the 200-500Hz range is enhanced (by 3-5dB).

[0068] Results: Noise suppression rate increased from 60% to 85%, and target speech signal-to-noise ratio (SNR) increased by 10dB.

[0069] When processing multi-band noise in combination: When CNN detects mid-to-high frequency noise (such as the sound of plates clattering in a restaurant, with an energy percentage of 500Hz-5kHz) System synchronization adjustment - intermediate frequency (500Hz-3kHz), set bandpass gain. Enhance the overtone range of human voices.

[0070] High frequency (>5kHz): Attenuation coefficient To suppress harsh noise.

[0071] The filter coefficient update rate has been increased to 200 times / second to ensure rapid response to sudden noise.

[0072] For the target speech signal, a dynamic range compression algorithm is used to dynamically adjust the compression ratio based on the instantaneous intensity of the speech signal. When the speech intensity is weak, the compression ratio is increased to make the weak speech clearer and more intelligible; when the speech intensity is strong, the compression ratio is appropriately reduced to avoid sound distortion.

[0073] In practical testing, this module was applied to multi-person conference scenarios. When up to 30 participants spoke simultaneously, users wearing this system were able to clearly distinguish the voices of speakers 3 meters away, with speech recognition accuracy increasing from 55% in traditional systems to 92%. In a restaurant environment with background noise reaching 75 decibels, users could still clearly hear whispers within 2 meters, with recognition accuracy increasing from 35% to 78%, significantly enhancing the system's adaptability to complex acoustic environments.

[0074] Example 3: Construction of a Low-Latency High-Fidelity Processing Module This embodiment aims to build a low-latency, high-fidelity processing module, which solves the problems of signal processing latency and audio quality loss through an innovative parallel processing architecture and advanced encoding and decoding technology.

[0075] This module employs a dual-core parallel processing architecture, with independent left and right ear signal processing units. Analog audio signals acquired by the microphone are simultaneously converted from digital to digital at a sampling frequency of 48kHz and 24-bit quantization precision. The converted digital signals are then fed into their respective processing units, where noise reduction, gain adjustment, and equalization operations are performed concurrently. To achieve efficient noise reduction, an optimized Wiener filtering algorithm is used. This algorithm is based on the minimum mean square error criterion and dynamically adjusts the filter coefficients by estimating the noise power spectrum in real time. In practical applications, the system updates the noise power spectrum estimate every 10 milliseconds and adjusts the filter coefficients based on the estimation results, effectively suppressing noise while keeping the noise reduction algorithm processing time within 3 milliseconds.

[0076] The encoding / decoding unit employs lossless audio coding technology based on wavelet transform. During encoding, the digital signal is first subjected to Discrete Wavelet Transform (DWT), decomposing it into different frequency sub-bands, including low-frequency approximation sub-bands and high-frequency detail sub-bands. For the low-frequency approximation sub-bands, which contain the main energy and key information of the signal, a higher-precision quantization method is used; for the high-frequency detail sub-bands, based on the masking effect of human hearing, the quantization precision is appropriately reduced while ensuring sound quality. The quantized sub-band data undergoes entropy coding to remove data redundancy, forming compressed encoded data. The decoding process is the reverse of encoding, accurately recovering the original signal through inverse entropy coding, inverse quantization, and inverse wavelet transform (IDWT). Testing shows that this encoding / decoding technology achieves an encoding / decoding latency of only 2 milliseconds while maintaining lossless sound quality.

[0077] In the signal output stage, the digital signal undergoes a 24-bit D / A conversion and is then amplified by a high-efficiency Class-D digital power amplifier. The Class-D power amplifier uses pulse-width modulation (PWM) technology to convert the input audio signal into a high-frequency pulse signal, which is then restored to an analog audio signal through a filtering circuit. Its efficiency reaches up to 85%, significantly reducing power consumption and heat generation compared to traditional linear power amplifiers.

[0078] Through actual audio playback tests, when playing high-resolution music with a sampling rate of 192kHz and a bit depth of 24bit, users can clearly distinguish the location and details of instruments, such as the friction sound of the violin bow and the sound of the piano keys. The perceptual deviation of music rhythm is less than 0.5%, and the subjective sound quality score reaches 9.2 out of 10. It achieves high-fidelity sound quality output while ensuring low latency, bringing users an excellent listening experience.

[0079] Example 4: Implementation of the Intelligent Personalized Debugging Module This embodiment details how the intelligent personalized debugging module achieves convenient and accurate personalized debugging through a mobile terminal APP and machine learning algorithms.

[0080] Upon first use, users open the mobile app and enter the hearing test interface. The app plays pure tone signals of different frequencies and intensities through headphones. Users provide feedback on the sounds they hear through the app, such as clicking the "heard" or "not heard" button. Based on the user's feedback data and a preset hearing test algorithm, the app calculates the user's binaural hearing curve, including the hearing threshold at each frequency point. The app also offers a variety of everyday life scenarios as options, such as daily conversation, watching TV, outdoor walks, and attending meetings. Users can select based on their primary activity scenarios, or multiple scenarios can be selected.

[0081] The app sends hearing profile data and selected living scenarios to the hearing aid's built-in control unit. The control unit contains a tuning model based on the Support Vector Regression (SVR) algorithm, trained using tuning data from 50,000 users with varying degrees of hearing loss. Taking the hearing profile and living scenario information as input, the model outputs personalized gain allocation and volume balancing schemes. For example, for users with severe high-frequency hearing loss who are frequently in noisy outdoor environments, the model automatically increases high-frequency gain and enhances noise suppression.

[0082] The app also supports manual parameter fine-tuning. Users can adjust parameters such as the gain curve slope, noise reduction intensity, and volume balance ratio through interactive methods like sliders and input boxes. During parameter adjustments, the app sends the adjusted parameters to the hearing aid in real time and provides built-in voice prompts, allowing users to immediately experience the improved sound effect. Simultaneously, the app's built-in sensors monitor the user's environment in real time, such as detecting ambient noise intensity via the phone's microphone and determining the user's movement status via an accelerometer. When the ambient noise intensity exceeds a certain threshold, the system automatically optimizes the tuning scheme to enhance noise suppression; when the app detects user movement, it appropriately adjusts the dynamic range of the sound to make the sound clearer and more stable.

[0083] Furthermore, the system possesses automatic learning and optimization capabilities. By recording data such as users' manual adjustments, usage time, and environmental information, it analyzes user preferences and habits. For example, if a user repeatedly adjusts a parameter in the same direction under a specific environment, the system will automatically incorporate that adjustment into subsequent default settings. Actual user testing shows that 95% of users can complete personalized adjustments via the app within 10 minutes, and the satisfaction rate with the adjusted hearing aids reaches 90% in different scenarios, significantly lowering the barrier to personalized adjustments.

[0084] Example 5: Design of a Durable Battery Life Module This embodiment mainly introduces the specific design and implementation of the durable battery life module in improving the protection performance and battery life of the device.

[0085] In terms of hardware protection, a nano-coating waterproof and dustproof technology is employed. First, the hearing aid shell undergoes surface pretreatment, using plasma cleaning technology to remove surface impurities and oil, improving the shell's hydrophilicity and adhesion. Then, a 50-nanometer-thick silica-carbon nanocomposite coating is applied to the shell surface using physical vapor deposition (PVD). This coating features a unique nanoscale porous structure with a surface contact angle greater than 150 degrees, exhibiting superhydrophobic properties while also repelling dust and other particles, achieving a superoleophobic effect. Furthermore, silicone rubber sealing rings are used to seal the joints between shell components, such as the battery compartment cover and charging port. These sealing rings employ a custom trapezoidal cross-section design, undergoing pre-compression during assembly to generate elastic deformation, forming a tight seal and achieving an overall IP67 protection rating for the hearing aid. Testing showed that after completely immersing the hearing aid in 1 meter of water for 30 minutes, the internal circuitry remained unaffected; and after 8 hours of continuous use in a dusty environment, no dust entered, effectively protecting the internal precision circuitry.

[0086] In terms of battery technology and battery life, a novel lithium-sulfur battery is adopted. The positive electrode material of the lithium-sulfur battery uses a nanoscale sulfur-carbon composite material. Through a special preparation process, sulfur is uniformly dispersed in a carbon nanotube network, improving the battery's charge-discharge performance and cycle stability. Its theoretical energy density reaches 420Wh / kg, which is 3.5 times that of traditional zinc-air batteries. To better manage battery power consumption, an intelligent power management algorithm was developed. This algorithm monitors the operating status of each module in the device in real time. When the Bluetooth connection is idle for more than 1 minute, it automatically switches the Bluetooth module from normal operation mode to low-power sleep mode. At this time, the power consumption of the Bluetooth module is reduced from 10mW to 1mW, a reduction of 90%. When the dual-ear signal processing load is low, such as when the user is in a quiet environment and there is no voice signal input, the processor automatically reduces the operation frequency from a maximum of 1GHz to 200MHz, and the power consumption is reduced by 60%. At the same time, the system has a battery health status monitoring function. By monitoring parameters such as battery voltage, charge-discharge current, and temperature, it uses equivalent circuit models and machine learning algorithms to predict the remaining battery capacity and remaining lifespan. When the battery level drops below 20%, the app will send a notification to the user and recommend nearby charging locations. Testing showed that with Bluetooth enabled and 8 hours of daily use, the hearing aid's battery life increased from 8 hours to 24 hours, meeting the user's all-day usage needs.

[0087] Example 6: Application of the Collaborative Adaptation Module This embodiment focuses on illustrating the adaptation scheme and application effect of the collaborative adaptation module for different special hearing loss groups.

[0088] For patients with profound hearing loss, the co-fit module features a dedicated cochlear implant co-working interface. This interface employs a high-speed wireless transmission protocol, operating in the 2.4-2.4835GHz frequency band, with a data transmission rate of 10Mbps. The hearing aid and cochlear implant communicate bidirectionally via this interface. The hearing aid encodes the processed sound signal (after noise reduction and enhancement) using a customized low-latency encoding algorithm, compressing the signal before wireless transmission to the cochlear implant. The entire transmission latency is controlled to within 5 milliseconds. Simultaneously, the cochlear implant feeds back its operational status information, such as electrode stimulation intensity and battery level, to the hearing aid. Based on this feedback, the hearing aid adjusts its signal processing strategy; for example, when the cochlear implant battery is low, the hearing aid appropriately reduces its output power to extend the cochlear implant's usage time. In practical applications, 20 patients with profound hearing loss who had received cochlear implants were selected for testing. After one month of use, the patients' speech recognition rate in quiet environments increased from 65% to 82% after cochlear implantation, and the recognition rate in noisy environments increased from 40% to 60%, significantly improving the patients' auditory experience.

[0089] For patients with central auditory processing disorders, a mobile app has been developed with a wealth of auxiliary training functions. In the sound localization training game, the app randomly plays sounds from different directions (front, back, left, right, upper left, upper right, etc., eight directions in total). Patients need to determine the direction of the sound source by operating buttons on their hearing aids or through the app interface. The system provides immediate feedback based on the patient's judgment: a green checkmark and points are awarded for correct judgment; a red cross and the correct direction are indicated for incorrect judgment. The speech recognition game plays speech signals with varying degrees of ambiguity, such as added noise, altered speech speed, and adjusted pitch. Patients need to recognize the speech content within a specified time. The system scores patients based on their recognition accuracy and reaction time, gradually increasing the degree of ambiguity to train their speech recognition abilities. After three months of training, 30 patients with central auditory processing disorders were tested. The patients' speech comprehension improved by an average of 32%, and their fluency in daily conversations significantly improved.

[0090] For users with asymmetric hearing loss, a personalized binaural balance algorithm model is established. First, using professional hearing testing equipment, the user's precise binaural hearing curves are obtained, and the difference in hearing loss between the two ears at various frequency points is calculated. Based on this difference, the amplification and frequency response of the sound in both ears are dynamically adjusted. For example, for a user with severe hearing loss in the left ear and mild hearing loss in the right ear, at a frequency of 1000Hz, the hearing loss in the left ear is 70dB and in the right ear is 20dB. The algorithm will set the amplification of the left ear to be 50dB higher than that of the right ear, while simultaneously providing an additional 3dB compensation to the left ear in the low-frequency range (20-500Hz). Furthermore, the algorithm will adaptively adjust based on real-time user feedback. For instance, if the user reports through the app that the sound in the left ear is still not clear enough, the system will further fine-tune the amplification and frequency response curve of the left ear. After testing on 50 users with asymmetric hearing loss, 90% of the users reported a significant improvement in binaural sound balance and a 45% increase in speech comprehension after wearing the device, effectively solving the problem of difficulty in fitting asymmetric hearing loss.

[0091] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of protection claimed by the present invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A binaural neural synchronous hearing aid system, characterized in that include: A high-precision synchronization module is used to optimize the Bluetooth transmission protocol and employ a dynamic time calibration algorithm to achieve high-precision synchronization of binaural signals. The intelligent acoustic processing module combines deep learning and adaptive filtering algorithms for noise suppression and speech enhancement. The low-latency, high-fidelity processing module reduces signal processing latency and ensures sound quality through a parallel processing architecture and efficient encoding / decoding technology. The intelligent personalized debugging module utilizes mobile terminal APP and machine learning algorithms to achieve personalized debugging. The durable battery module uses nano-coating waterproof and dustproof technology and high-energy-density batteries to improve the device's durability and battery life. The collaborative adaptation module provides adaptation solutions for users with different types of hearing loss.

2. The binaural neural synchronous hearing aid system according to claim 1, characterized in that: The high-precision synchronization module optimizes the data packet structure of the Bluetooth transmission protocol, adopts time division multiplexing technology, and the dynamic time calibration algorithm is based on Kalman filtering principle, and the formula 、 Real-time monitoring and adjusting the left and right ear signal transmission time; Where X(k) is the state vector, A is the state transition matrix, B is the control matrix, U(k) is the control vector, W(k) is the process noise, Z(k) is the measurement vector, H is the measurement matrix, and V(k) is the measurement noise.

3. The binaural neural synchronous hearing aid system according to claim 1, characterized in that: The intelligent acoustic processing module is constructed based on a convolutional neural network and an adaptive filtering algorithm, the convolutional neural network contains no less than 5 convolutional layers, 3 pooling layers and 2 fully connected layers, sound characteristics are recognized by training a large amount of acoustic data, and the adaptive filtering algorithm adopts a low-frequency promotion filtering formula Process low-frequency sound.

4. The binaural neural synchronous hearing aid system of claim 1, wherein: The low-latency high-fidelity processing module adopts a parallel processing architecture, with the left and right ear signal processing units sampling synchronously at a sampling frequency of no less than 48kHz and a quantization bit depth of no less than 24 bits. The encoding and decoding unit adopts lossless audio coding technology based on wavelet transform, with an encoding and decoding delay of no more than 2 milliseconds. The noise reduction algorithm adopts an optimized Wiener filtering algorithm.

5. The binaural neural synchronous hearing aid system of claim 1, wherein: The intelligent personalized tuning module's mobile terminal APP allows users to input hearing curve data and select life scene modes. It generates personalized solutions using a tuning model based on support vector regression algorithm, while also supporting manual fine-tuning of parameters and automatically optimizing the solution based on user auditory feedback, thus shortening the tuning time.

6. The binaural neural synchronous hearing aid system of claim 1, wherein: In the durable battery life module, the nano-coating waterproof and dustproof technology forms a superhydrophobic and superoleophobic coating with a thickness of 30-80 nanometers on the surface of the shell, which, together with the sealing structure design, achieves an IP67 protection level. It uses a lithium-sulfur battery with an energy density of no less than 350Wh / kg and uses an intelligent power management algorithm to monitor the operating status of each module in real time. When the Bluetooth connection is idle for more than a preset time, it automatically switches to a low-power sleep mode to reduce the power consumption of the Bluetooth module and extend the battery life of the device when the Bluetooth function is turned on.

7. The binaural neuro-synchronous hearing aid system of claim 1, wherein: The collaborative adaptation module is designed for patients with profound hearing loss. It has a dedicated cochlear implant collaborative working interface, developed a unified synchronization protocol, controls the signal transmission delay between the hearing aid and the cochlear implant to within 5 milliseconds, and realizes two-way information interaction. For patients with central auditory processing disorders, the mobile app provides at least two types of auditory training games, including sound localization and speech recognition. For users with asymmetric hearing loss, a personalized binaural balance algorithm model based on the difference in binaural hearing curves is established, which can dynamically adjust the difference in sound amplification between the two sides between 15-25dB and the low-frequency compensation between 2-5dB.

8. The binaural nerve-synchronous hearing aid system of claim 1, wherein: During the system initialization phase, the high-precision synchronization module obtains the initial delay time of the left and right ear signal transmission by sending a test signal, which is used as the initial value of the state vector of the dynamic time calibration algorithm. During operation, the actual arrival time of the left and right ear signals is collected at fixed time intervals of 5-15 milliseconds for algorithm iterative calculation.

9. The binaural nerve-synchronous hearing aid system of claim 1, wherein: The intelligent acoustic processing module preprocesses the acquired sound signals, including normalization to limit the signal amplitude to between -1 and 1, and frame processing with each frame lasting 10-30 milliseconds. In the noise reduction process, it automatically adjusts the adaptive filter coefficients based on the noise intensity and spectral characteristics determined by the CNN model, with an adjustment step size of 0.01-0.

1.

10. The binaural nerve-synchronous hearing aid system of claim 1, wherein: The output unit of the low-latency high-fidelity processing module has a D / A conversion accuracy of no less than 24 bits, and the power amplifier adopts a Class-D digital power amplifier, which is present throughout the entire signal processing link.