An Adaptive Intelligent Noise Reduction Method for Audio System Transmission

Abstract

The present invention relates to the technical field of digital information transmission, and specifically to an adaptive intelligent noise reduction method for audio system transmission, including: First, obtain the transmission channel state information and design dynamic pilots to extract key channel parameters. A dual-loop processing model is adopted, where the inner loop performs time delay compensation through the LMS algorithm, and the outer loop uses Kalman filtering to estimate the channel gain and updates the channel impulse response to optimize the channel state recovery. Then, extract the audio features, and perform speech noise reduction by combining the improved NMF model and the noise dictionary, and at the same time optimize the noise dictionary in real time through the dynamic dictionary mechanism. The noise-reduced speech uses the PTP protocol for clock synchronization, and executes NACK retransmission requests or decodes through FEC according to the packet loss rate detection results, improving the accuracy and reliability of audio system transmission.

Description

Technical Field

[0001] The present invention relates to the technical field of digital information transmission, and particularly to an adaptive intelligent noise reduction method for audio system transmission. Background Art

[0002] In the process of data transmission and signal processing, especially in large-scale space application scenarios, the stability, anti-interference ability of audio data in network transmission, and the latency problem of end-to-end synchronization processing have a significant impact on the quality of audio communication. Therefore, it is of great practical significance to develop an audio data transmission method based on adaptive signal processing and intelligent noise reduction.

[0003] In a large-scale space environment, existing adaptive transmission schemes are usually based on preset channel state assumptions, making it difficult to cope with sudden changes in acoustic parameters caused by personnel movement and unable to effectively handle the multipath time-varying effect caused by building structures. In addition, in a distributed audio data transmission system, there are multiple noise interferences from different sources, such as air-conditioning fans, electromagnetic interference, keyboard keystrokes, and network jitter. The spectral characteristics of these interferences may have a large overlap. When multiple noise sources interfere with each other, it is difficult for these adaptive transmission schemes to effectively model the spatial distribution of noise, thus affecting the accuracy of noise reduction and further leading to an increase in the bit error rate of data transmission. Finally, in an audio data transmission network with multi-device collaboration, it is usually necessary to perform timing synchronization and data fusion on multiple audio data streams. However, these adaptive transmission schemes adopt a centralized data processing architecture and rely on a central server for data aggregation and signal synchronization, resulting in an increase in the latency of audio data transmission. When the network load is high, problems such as packet loss and computing resource bottlenecks may occur, thus affecting the real-time performance and synchronous noise reduction effect of audio data.

[0004] In order to improve the accuracy of audio system transmission, an adaptive intelligent noise reduction method for audio system transmission is proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide an adaptive intelligent noise reduction method for audio system transmission. By obtaining the transmission channel state information and designing dynamic pilots to extract key channel parameters. A double-loop processing model is adopted, where the inner loop performs delay compensation through the LMS algorithm, and the outer loop uses Kalman filtering to estimate the channel gain and updates the channel impulse response to optimize the channel state recovery. Then, audio features are extracted, and speech noise reduction is performed by combining an improved NMF model and a noise dictionary, while the noise dictionary is optimized in real time through a dynamic dictionary mechanism. The noise-reduced speech is synchronized using the PTP protocol, and NACK retransmission requests are executed or decoded through FEC according to the packet loss rate detection results, improving the accuracy and reliability of audio system transmission.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An adaptive intelligent noise reduction method for audio system transmission, comprising:

[0008] Obtaining transmission channel state information;

[0009] Designing a dynamic pilot for extracting a pilot part according to channel parameters and environmental parameters of the transmission channel state information;

[0010] Establishing a double-loop processing model, including: inner-loop processing for time-domain alignment of the pilot part and a local reference sequence using the LMS algorithm to output delay compensation; outer-loop processing for estimating channel gain using a Kalman filter; updating the channel impulse response using the delay compensation and the channel gain to restore the transmission channel state information;

[0011] Extracting audio features of the transmission channel state information, using an improved NMF model and a noise dictionary to decompose and denoise speech and noise basis; updating the noise dictionary using a dynamic dictionary mechanism;

[0012] Performing clock synchronization marking on the denoised speech using the PTP protocol and detecting the packet loss rate. If the packet loss rate is greater than a preset condition, sending a NACK retransmission request; otherwise, decoding the denoised speech using FEC according to the dynamic redundancy mechanism.

[0013] Further, the transmission channel state information includes:

[0014] The channel parameters, including multi-channel speech signals collected by a microphone array;

[0015] Wireless network channel parameters, including network bandwidth, the packet loss rate, and delay jitter;

[0016] The environmental parameters, including three-dimensional spatial dimensions, sound source position, reflector material, and number of people.

[0017] Further, the design process of the dynamic pilot includes:

[0018] Defining the pilot part using a dynamic adjustment strategy, expressed as:

[0019] ;

[0020] Wherein, is the pilot part, is the pilot amplitude adjusted according to the environmental parameters, is the cosine function, is the pilot frequency, is the time variable;

[0021] Extract the pilot part according to the pilot part , expressed as:

[0022] ;

[0023] wherein, is the channel impulse response, is the additive noise, is the convolution operation, is the discrete time index.

[0024] Furthermore, the extraction process of the audio features includes:

[0025] Perform multi-channel frequency band decomposition on the transmission channel state information using a Gammatone filter bank to obtain audio sub-band signals;

[0026] Calculate the spatial covariance matrix for the audio sub-band signals to analyze the spatial distribution characteristics;

[0027] Perform MFCC feature extraction on the audio sub-band signals and combine dynamic differential coefficients to obtain spectral temporal characteristics.

[0028] Furthermore, the optimization objective function of the improved NMF model includes:

[0029] ;

[0030] wherein, is the audio feature matrix, is the basis matrix, is the weight matrix, is the temporal continuity constraint term, is the basis sparsity constraint term, and are the adjustment weights, is the Frobenius norm, is the weight matrix of the k-th frame, is the weight matrix of the (k - 1)-th frame.

[0031] Furthermore, the construction process of the noise dictionary includes:

[0032] Collect noise samples;

[0033] For each type of the noise samples, extract the noise basis using the improved NMF model;

[0034] For the noise basis, screen the basis matrices that meet the energy concentration threshold and the time continuity threshold to generate the noise dictionary.

[0035] Furthermore, the update process of the noise dictionary includes:

[0036] Calculate the matching error between the noise basis and the noise dictionary , expressed as:

[0037] ;

[0038] Where is the audio feature, is the existing noise basis, is the weight matrix corresponding to the existing noise basis, is the Frobenius norm;

[0039] If it is detected that the matching error exceeds the set threshold and the duration exceeds N seconds, then calculate the orthogonal projection residual of the new noise basis , expressed as:

[0040] ;

[0041] Where is the existing noise basis, is the pseudoinverse of;

[0042] If is greater than the preset residual threshold, add to the noise dictionary and optimize the noise dictionary through atomic pruning and Gram - Schmidt orthonormalization.

[0043] Furthermore, the dynamic redundancy mechanism includes:

[0044] Adjust the redundancy according to the packet loss rate;

[0045] According to the redundancy, use the FEC to optimize the audio transmission quality for coding parameters under different network conditions.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] 1. Through the dynamic pilot extraction mechanism, the present invention adaptively extracts the pilot part according to the time - varying characteristics of the channel state information, thereby improving the accuracy of channel estimation. In addition, by constructing a double - loop processing model, where the inner loop processing uses the LMS algorithm for time - domain alignment to compensate for the delay error, and the outer loop processing uses the Kalman filter to estimate the channel gain in real - time, and finally combines the delay compensation and gain information to optimize the channel impulse response, thus effectively improving the accuracy of channel state recovery and enhancing the adaptability and robustness of the audio system in complex environments.

[0048] 2. The present invention improves the accuracy of audio feature extraction through Gammatone filters, spatial covariance matrices, and MFCC feature extraction. It combines an improved NMF model with a noise dictionary for speech decomposition and noise reduction, achieving efficient audio signal enhancement. Additionally, by using a dynamic dictionary update mechanism, it can detect noise changes in real time and dynamically expand the noise dictionary, ensuring adaptive noise reduction capabilities in complex noise environments, thereby improving the clarity and separability of audio signals and enhancing the accuracy of audio recognition.

[0049] 3. The present invention integrates the PTP protocol synchronization and an adaptive packet loss recovery mechanism. During the transmission of noise-reduced speech data, high-precision clock synchronization is achieved through the PTP protocol, effectively reducing transmission delay errors. At the same time, the packet loss rate is monitored in real time. If the packet loss rate exceeds the threshold, a NACK request is triggered for data retransmission; otherwise, based on the dynamic redundancy mechanism, FEC error correction decoding is used to improve the stability of speech data transmission and enhance the communication reliability and noise reduction accuracy of the audio system in complex network environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is a schematic flowchart of an adaptive intelligent noise reduction method for audio system transmission provided by the present invention;

[0051] Figure 2 is a schematic flowchart of the process for decomposing and denoising speech and noise bases provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0053] Please refer to Figures 1 to 2 , the present invention provides an adaptive intelligent noise reduction method for audio system transmission, and the technical solutions are as follows:

[0054] In large-scale spaces and distributed audio transmission systems, audio signal transmission faces many challenges. For example, dynamic acoustic environments and multipath effects may make traditional solutions difficult to adapt to, resulting in unstable signals; in addition, multi-source noise interference is difficult to effectively distinguish and suppress, affecting transmission quality; finally, a centralized processing architecture is prone to bottlenecks under high loads, increasing latency and making it difficult to meet personalized noise reduction requirements; existing noise reduction solutions lack adaptive optimization capabilities and are difficult to cope with complex and changing environment and channel conditions.

[0055] To improve the accuracy of audio system transmission in a large space, the following is the first embodiment:

[0056] Figure 1 It is a schematic flow chart of an adaptive intelligent noise reduction method for audio system transmission provided by the present invention.

[0057] As Figure 1 shown, an adaptive intelligent noise reduction method for audio system transmission includes:

[0058] Step 1: Obtain the transmission channel status information;

[0059] Furthermore, the transmission channel status information includes:

[0060] Channel parameters, including multi-channel voice signals collected by a microphone array;

[0061] Wireless network channel parameters, including network bandwidth and packet loss rate, for adjusting audio coding and error correction mechanisms to ensure data reliability; delay jitter, for monitoring network transmission jitter and optimizing timing synchronization strategies;

[0062] Environmental parameters, including three-dimensional space dimensions, sound source position, reflecting surface material, and number of people.

[0063] In addition, it may also include device status information, such as microphone gain and frequency response curve, for adaptively adjusting signal processing parameters. By obtaining this transmission channel status information, the audio signal processing strategy can be adjusted in real time, the data transmission quality can be optimized, and the anti-interference ability can be enhanced, thereby improving the accuracy of audio system transmission.

[0064] Step 2: Design a dynamic pilot for extracting the pilot part according to the channel parameters and environmental parameters of the transmission channel status information.

[0065] Furthermore, the design process of the dynamic pilot includes:

[0066] Define the pilot part using a dynamic adjustment strategy, expressed as:

[0067] ;

[0068] Wherein, is the pilot part, is the pilot amplitude adjusted with the environmental parameters, is the cosine function, is the pilot frequency, is the time variable;

[0069] Extract the pilot part according to the pilot part , expressed as:

[0070] ;

[0071] Among them, is the channel impulse response, is the additive noise, is the convolution operation, is the discrete-time index.

[0072] By dynamically adjusting the pilot transmission interval when the personnel density changes, the high-efficiency channel tracking ability can be maintained, the channel changes can be captured more timely, and the accuracy of audio transmission can be improved.

[0073] Step three: Establish a double-loop processing model, including: inner-loop processing, which is used to align the time domain of the pilot part with the local reference sequence using the LMS (Least Mean Squares) algorithm and output the delay compensation; outer-loop processing, which is used to estimate the channel gain using a Kalman filter; use the delay compensation and the channel gain to update the channel impulse response and restore the transmission channel state information.

[0074] Specifically, in this embodiment, in the inner-loop processing, the local reference sequence is used for delay compensation. The transmitting end pre-generates a fixed pilot part and inserts it into the audio data stream during transmission. The receiving end stores the same local reference sequence for comparing the received pilot part, calculating the delay error, and adjusting the received signal through the least mean square error algorithm to align it with the local reference sequence on the time axis, eliminating the delay drift during transmission. In the outer-loop processing, the Kalman filter can estimate the channel gain in real time to reflect the dynamic changes of the channel and optimize the signal quality under the multipath effect. Then, through joint processing, combining the delay compensation and the channel gain to update the channel impulse response , thereby reducing the multipath interference and improving the accuracy of signal recovery.

[0075] Step four: Figure 2 is the schematic diagram of the process for decomposing and reducing noise of speech and noise basis provided by the present invention. As Figure 2 shown, extract the spatial distribution characteristics and spectral time-series characteristics of the transmission channel state information, and fuse them to form a joint feature matrix to obtain audio features. Use the improved NMF (Non-negative Matrix Factorization) model and the noise dictionary to decompose and reduce noise of speech and noise basis for the audio features; use the dynamic dictionary mechanism to update the noise dictionary.

[0076] Furthermore, the extraction process of the audio features includes:

[0077] The multi-channel frequency band decomposition of the transmission channel state information is performed using a Gammatone filter bank to obtain audio sub-band signals, expressed as:

[0078] ;

[0079] wherein, is the audio sub-band signal, is the transmission channel state information, is the Gammatone filter with a center frequency of ;

[0080] Among them, 32 filters are adopted in this embodiment, and the center frequency range is 50 Hz to 8000 Hz.

[0081] Calculate the spatial covariance matrix of the audio sub-band signal , for analyzing the spatial distribution characteristics, expressed as:

[0082] ;

[0083] wherein, is the window length, is the multi-channel audio sub-band signal matrix of the k-th frame, is the conjugate transpose of the matrix;

[0084] Among them, in this embodiment, the covariance matrix within a 50 ms time window is calculated, and the maximum eigenvalue and its corresponding eigenvector are extracted for enhancing the target speech signal.

[0085] MFCC (Mel-Frequency Cepstral Coefficients) feature extraction is performed on the audio sub-band signal, and spectral temporal features are obtained by combining dynamic differential coefficients. Specifically, it includes: calculating the short-time power spectrum of the audio sub-band signal using the short-time Fourier transform, obtaining the Mel spectrum by applying the Mel filter bank, taking the logarithm of the Mel spectrum, and then performing the discrete cosine transform to obtain MFCC, and calculating the dynamic differential coefficients based on MFCC to form spectral temporal features.

[0086] By using Gammatone filtering to enhance the key band signals of the transmission channel state information, the noise interference is effectively reduced. At the same time, the covariance matrix is calculated to analyze the sound source distribution, improving the spatial information extraction ability. On this basis, MFCC and its dynamic differences are extracted to further enhance the robustness of the speech features, thereby improving the accuracy of audio transmission.

[0087] Furthermore, the optimization objective function of the improved NMF model includes:

[0088] ;

[0089] Among them, is the audio feature matrix, where each column represents a time frame and each row represents a feature dimension. is the basis matrix, where each column represents a fundamental frequency mode. is the weight matrix, where each column represents the activation degree of each basis at each time frame. is the time-domain continuity constraint term, which ensures the smooth time variation of the speech signal and reduces the mutations caused by noise. is the basis sparsity constraint term, which ensures that the speech and noise bases do not overlap as much as possible and improves the separability of the speech. and are the adjustment weights. is the Frobenius norm.

[0090] Among them, is the weight matrix of the k-th frame. is the weight matrix of the (k - 1)-th frame.

[0091] Specifically, first input the pre-trained noise dictionary , the basis matrix of the target speech (obtained from a clean speech library) and the audio features. Then, initialize the basis matrix and the weight matrix , being random values. Then, use gradient descent to update the parameters of and in the optimization objective function. Determine which bases belong to speech and which belong to noise through the decomposed weight matrix , retain the related part, and remove the related to the noise bases to obtain the denoised audio signal.

[0092] The improved NMF model introduces a time-domain continuity constraint to reduce the mutations of the speech signal and improve the smoothness of the speech. At the same time, the basis sparsity constraint enhances the distinguishability between the speech and noise bases and prevents noise from mixing into the speech. In addition, the use of the noise dictionary is added, enabling the method to adapt to various background noises and improving the accuracy of audio transmission.

[0093] Furthermore, the construction process of the noise dictionary includes:

[0094] Collect noise samples;

[0095] For each type of the noise samples, use the improved NMF model to extract the noise bases;

[0096] For the noise floor, select basis matrices that meet the energy concentration threshold and the time continuity threshold to generate the noise dictionary.

[0097] Specifically, in this embodiment, first extract common noise categories from the public dataset, such as keyboard tapping and air conditioner operation sounds. The sampling rate is set to 16 kHz, and the length is intercepted for 10 s to ensure consistency. Convert the noise samples into time-frequency features, and use the improved NMF model to extract the noise basis , , and are different noise basis patterns.

[0098] Calculate the energy concentration, expressed as:

[0099] ;

[0100] where is the energy concentration of the a-th basis, is the element in the a-th row and b-th column of the basis matrix, is the element in the b-th row and a-th column of the weight matrix;

[0101] Set the energy concentration threshold. If is greater than the energy concentration threshold, then this basis is retained, otherwise it is discarded.

[0102] Calculate the time continuity, expressed as:

[0103] ;

[0104] where is the time continuity of the a-th basis, is the activation value of the weight matrix at b;

[0105] Set the time continuity threshold. If is less than the time continuity threshold, then this basis is retained, otherwise it is discarded.

[0106] Take the intersection of the two thresholds to determine the final noise basis. Construct and pre-train the noise dictionary, and directly apply the known noise patterns during noise reduction. The pre-trained dictionary enables denoising only by sparse representation matching, with a lower computational cost than online training, and provides an accurate noise basis for audio system denoising.

[0107] Furthermore, the update process of the noise dictionary includes:

[0108] Calculate the matching error between the noise basis and the noise dictionary , expressed as:

[0109] ;

[0110] Among them, is the audio feature, is the existing noise floor, is the weight matrix corresponding to the existing noise floor, is the Frobenius norm;

[0111] If it is detected that the matching error exceeds the set threshold and the duration exceeds N seconds, then calculate the orthogonal projection residual of , which is expressed as:

[0112] ;

[0113] Among them, is the existing noise floor, is the pseudoinverse of;

[0114] If is greater than the preset residual threshold, add to the noise dictionary and optimize the noise dictionary through atomic pruning and Gram - Schmidt orthogonalization.

[0115] Atomic pruning can delete redundant or weakly activated bases to prevent the dictionary from being too large, and Gram - Schmidt orthogonalization ensures that the dictionary bases are approximately orthogonal, enhancing the noise separation effect. The updated noise dictionary can more accurately represent the new noise, improve the extraction quality of the audio system, and thus ensure the noise reduction effect in a large - space environment.

[0116] Step Five: Perform clock synchronization marking on the noise - reduced speech using the PTP (Precision Time Protocol) protocol and detect the packet loss rate. If the packet loss rate is greater than the preset condition, send a NACK (Negative Acknowledgment) re - transmission request; otherwise, decode the noise - reduced speech using FEC (Forward Error Correction) according to the dynamic redundancy mechanism.

[0117] Furthermore, the dynamic redundancy mechanism includes:

[0118] Adjust the redundancy according to the packet loss rate;

[0119] According to the redundancy, use the FEC to optimize the audio transmission quality for encoding parameters under different network conditions.

[0120] Specifically, to ensure the timing consistency and data integrity of audio data, during the transmission of the noise-reduced audio data, a timestamp based on the PTP protocol is added to each data packet to ensure the time synchronization of all receiving-end devices and avoid audio distortion caused by out-of-order data packets. The receiving end detects the packet loss rate. If the packet loss rate is greater than the preset threshold, an NACK request is sent to require the sending end to retransmit the lost voice data packets. If the packet loss rate is not greater than the preset threshold, FEC is used for decoding processing according to the dynamic redundancy mechanism. For example, assuming that the current network jitter is small and the packet loss rate is 1%, which is lower than the preset threshold of 5%, and the redundancy is set to 10%, some of the lost data packets can be recovered. Assuming that the current network jitter is large and the packet loss rate is 4%, which is still less than the preset threshold, the FEC mechanism adjusts the redundancy to 20% to enhance the error correction ability.

[0121] Compared with the FEC with fixed redundancy, the dynamic redundancy mechanism can adaptively adjust the redundancy according to the network conditions, which not only improves the transmission efficiency but also enhances the accuracy of audio transmission, providing a more reliable audio experience for users.

[0122] First, the pilot is extracted in real time through dynamic pilot design, improving the accuracy of channel information acquisition. Then, a double-loop processing model is adopted. The inner loop uses the LMS algorithm to align the signals and compensate for the delay, reducing audio distortion. The outer loop uses Kalman filtering to estimate the channel gain, enhancing the data recovery ability. Combining the delay compensation and the dynamic adjustment of the channel gain for the channel model can adapt to complex environments. Then, an improved NMF noise reduction model is designed, which adopts time-domain continuity and basis sparsity constraints to improve the voice noise separation ability. Combining the dynamic dictionary update mechanism, new bases are automatically added when the environmental noise changes, and the noise reduction effect can be stabilized. Finally, during transmission, a packet loss detection and retransmission mechanism is used, and the FEC error correction strength is adjusted with dynamic redundancy, balancing the bandwidth and error correction, thereby improving the accuracy of audio transmission.

[0123] Embodiment 2:

[0124] In modern intelligent conference rooms, since the conference room is usually a large enclosed space, its audio transmission system is easily affected by factors such as multipath effects, environmental noise, and personnel movement, resulting in dynamic changes in the transmission channel. Traditional noise reduction and signal processing methods are difficult to meet the actual needs. Taking the conference room of a certain technology company as an example, an adaptive intelligent noise reduction method for audio system transmission is proposed to achieve efficient noise reduction and channel adaptive processing, ensuring the accuracy of conference audio, specifically including:

[0125] Step 1: Obtain the transmission channel state information;

[0126] Furthermore, in this embodiment, the transmission channel state information includes:

[0127] Channel parameters, including: multi-channel voice signals collected by a microphone array, combined with multipath effect parameters (signal attenuation and time delay), echo path characteristics (calculating the room echo path model based on an echo cancellation algorithm), and signal-to-noise ratio information (for dynamic gain control), to analyze the impact of the environment on signal propagation.

[0128] Wireless network channel parameters, including: network bandwidth and packet loss rate, used to adjust audio coding and error correction mechanisms to ensure data reliability; delay jitter, calculating the arrival time difference between adjacent audio data packets based on the timestamps of the received audio data packets, and using a sliding window (window size 1s) to calculate the standard deviation to estimate the range of delay jitter, thereby optimizing the timing synchronization strategy.

[0129] Environmental parameters, including: three-dimensional spatial dimensions, set by the conference room structure parameters. Sound source location, calculating the sound source coordinates using the time difference of arrival (TDOA) algorithm to determine the sound source location; reflecting surface material, combining a pre-stored reflection coefficient table and determining the reflecting surface material according to the material properties of different areas of the conference room; number of people, using the YOLOv5 model to count the number of people in real time.

[0130] Step 2: Design a dynamic pilot, used to extract the pilot part according to the channel parameters and environmental parameters of the transmission channel state information. Because in a conference room environment, the propagation of sound is affected by the multipath effect, resulting in phase shift and amplitude attenuation when the signal arrives at the microphone after propagating through different paths. It is necessary to compensate for these distortions to improve audio quality.

[0131] Furthermore, the design process of the dynamic pilot includes:

[0132] Defining the pilot part using a dynamic adjustment strategy, adopting a periodic known signal, expressed as:

[0133] ;

[0134] where is the pilot part, is the pilot amplitude adjusted according to the environmental parameters, is the cosine function, is the pilot frequency, set to 500, is the time variable;

[0135] Extracting the pilot part according to the pilot part , expressed as:

[0136] ;

[0137] where is the channel impulse response, is the additive noise, is a convolution operation, is a discrete-time index.

[0138] Specifically, the meeting room size is 18m×12m×4m, the sound source position is that the speaker is 6m away from the microphone, the reflecting surface is a glass curtain wall (reflection coefficient 0.85), and the other walls are concrete. The number of people is initially 15 and increases to 25 during the meeting. Based on these data, the calculated data is shown in Table 1, and the personnel density changes from 0.07 person per square meter to 0.12 person per square meter. When there are 15 people, is 1.7, and when there are 25 people, is 2.2, ensuring that the pilot part can still be detected when the noise and reflection are enhanced. The pilot transmission interval threshold is set to 0.1, so the pilots are extracted at intervals of 50ms and 30ms respectively. The transmission interval is shortened from 50ms to 30ms, updating the channel state more frequently. Through dynamic adjustment, the pilots can more accurately estimate the phase shift and amplitude attenuation caused by the multipath effect and noise, thereby improving the audio accuracy.

[0139] Table 1 Dynamic Pilot Data

[0140]

[0141] Step 3: Establish a double-loop processing model, including: inner-loop processing, which is used to align the time domain of the pilot part with the local reference sequence using the LMS algorithm and output the delay compensation; outer-loop processing, which is used to estimate the channel gain using the Kalman filter; and updating the channel impulse response using the delay compensation and the channel gain to restore the transmission channel state information.

[0142] Specifically, as shown in Table 2, the test conditions are that the air conditioner is continuously turned on, there are sudden keyboard taps, and two people cross-walk at a speed of about 1m / s. The separate LMS algorithm and the double-loop processing model of the present invention are used for comparison,

[0143] Table 2 Effect Comparison

[0144]

[0145] Step 4: Extract the audio features of the transmission channel state information, use the improved NMF model and the noise dictionary to decompose the noise-reduced speech and the noise base; update the noise dictionary using the dynamic dictionary mechanism.

[0146] Step 5: Perform clock synchronization marking on the noise-reduced speech using the PTP protocol and detect the packet loss rate. If the packet loss rate is greater than the preset condition, send a NACK retransmission request; otherwise, decode the noise-reduced speech using FEC according to the dynamic redundancy mechanism.

[0147] Specifically, the deployment plan for a company's meeting room is shown in Table 3. For example, under the 5G network, the packet loss rate is relatively low. Therefore, 10% redundant coding is adopted, and a packet length of 1KB is used to optimize the transmission efficiency. Under this configuration, 3% of random packet losses can be effectively repaired, ensuring the stability of audio transmission. Compared with the traditional fixed redundancy scheme, the average decoding delay of the present invention is about 12ms, while the decoding delay of the traditional fixed redundancy scheme is about 16 seconds. The present invention improves the decoding speed and at the same time improves the accuracy of audio system transmission.

[0148] Table 3 Meeting Room Deployment Plan

[0149]

[0150] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An audio system transmission adaptive intelligent noise reduction method, characterized in that, Including: Obtain the transmission channel state information; Design a dynamic pilot, which is used to extract the pilot part according to the channel parameters and environmental parameters of the transmission channel state information; Establish a double-loop processing model, including: inner-loop processing, which is used to align the time domain of the pilot part with the local reference sequence using the LMS algorithm and output the delay compensation; outer-loop processing, which is used to estimate the channel gain using a Kalman filter; use the delay compensation and the channel gain to update the channel impulse response and recover the transmission channel state information; Extract the audio features of the recovered transmission channel state information, use the improved NMF model and the noise dictionary to decompose the noise-reduced speech and the noise base; use the dynamic dictionary mechanism to update the noise dictionary; Perform clock synchronization marking on the noise-reduced speech using the PTP protocol and detect the packet loss rate. If the packet loss rate is greater than the preset condition, send a NACK retransmission request; otherwise, decode the noise-reduced speech using FEC according to the dynamic redundancy mechanism.

2. The adaptive intelligent noise reduction method for audio system transmission according to claim 1, wherein The transmission channel state information includes: The channel parameters, including multi-channel speech signals collected by a microphone array; Wireless network channel parameters, including network bandwidth, the packet loss rate, and delay jitter; The environmental parameters, including the three-dimensional size of the space, the sound source position, the material of the reflecting surface, and the number of people.

3. An audio system transmission adaptive intelligent noise reduction method according to claim 1, characterized in that, The design process of the dynamic pilot includes: Use a dynamic adjustment strategy to define the pilot part, expressed as: ； wherein, is the defined pilot part, is the pilot amplitude adjusted according to the environmental parameter, is the cosine function, is the pilot frequency, is the time variable; According to the defined pilot part Extract the pilot part , expressed as: ； wherein, is the channel impulse response, is the additive noise, is the convolution operation, is the discrete-time index.

4. An audio system transmission adaptive intelligent noise reduction method according to claim 1, characterized in that, The extraction process of the audio features includes: Use a Gammatone filter bank to perform multi-channel band decomposition on the transmission channel state information to obtain audio sub-band signals; Calculate the spatial covariance matrix of the audio sub-band signals for analyzing the spatial distribution characteristics; Extract MFCC features from the audio sub-band signals and combine them with dynamic differential coefficients to obtain spectral temporal features.

5. An audio system transmission adaptive intelligent noise reduction method according to claim 1, characterized in that The optimization objective function of the improved NMF model includes: ； Among them, is the audio feature matrix, is the basis matrix, is the weight matrix, is the time-domain continuity constraint term, is the basis sparsity constraint term, and are the adjusted weights, is the square of the Frobenius norm, is the weight matrix of the k-th frame, is the weight matrix of the (k - 1)-th frame.

6. An audio system transmission adaptive intelligent noise reduction method according to claim 1, characterized in that The construction process of the noise dictionary includes: Collect noise samples; For each type of the noise samples, use the improved NMF model to extract the noise base; For the noise base, screen the base matrices that meet the energy concentration threshold and the time continuity threshold to generate the noise dictionary.

7. An audio system transmission adaptive intelligent noise reduction method according to claim 1, characterized in that The update process of the noise dictionary includes: Calculate the matching error between the noise floor and the noise dictionary , expressed as: ； Among them, is the audio feature, is the existing noise floor, is the weight matrix corresponding to the existing noise floor, is the Frobenius norm; If it is detected that the matching error exceeds a set threshold and the duration exceeds N seconds, then a new noise floor is calculated of the orthogonal projection residual , which is expressed as: ； Among them, is the existing noise floor, is the pseudo-inverse of; If is greater than a preset residual threshold, add to the noise dictionary, and optimize the noise dictionary through atomic pruning and Gram - Schmidt orthogonalization.

8. An audio system transmission adaptive intelligent noise reduction method according to claim 1, characterized in that, The dynamic redundancy mechanism includes: Adjust the redundancy according to the packet loss rate; According to the redundancy, use FEC to optimize the audio transmission quality for the coding parameters under different network conditions.

Images