Active noise reduction method and device based on SOC chip

Abstract

The invention relates to an active noise reduction method and device based on an SOC chip, and the method comprises the following steps: carrying out the blind source separation of a multi-channel noise signal, and obtaining a separated noise sub-signal; obtaining a target noise source and direction information corresponding to the target noise source, and performing direction focusing on the target noise source based on the separated noise sub-signals and the direction information through an acoustic positioning algorithm in a target SOC chip to obtain directional noise signals; performing time-frequency domain conjoint analysis on the directional noise signal through a target SOC chip to obtain time-frequency characteristic information of the directional noise signal; performing reverse noise generation on the directional noise signal based on the time-frequency characteristic information to obtain a reverse noise waveform; and performing phase inversion compensation on the directional noise signal based on the reverse noise waveform to obtain a denoised target signal, thereby solving the technical problem of how to accurately collect and separate multi-channel noise signals and reduce noise.

Description

Technical Field

[0001] The present invention relates to the technical field of SOC chips, and in particular to an active noise reduction method and device based on a SOC chip. Background Art

[0002] In modern society, as people's requirements for the quality of sound environment continue to increase, especially the growing demand for quietness in noisy environments, active noise reduction technology has gradually become a research hotspot. Traditional passive noise reduction methods, such as using earplugs or soundproofing materials, can reduce the impact of external noise to a certain extent, but the effect on low-frequency noise is limited, and they are large in size and not portable. Therefore, active noise reduction technology based on SOC (System on Chip) chips has received widespread attention as an emerging solution. It directly processes noise through electronic technology and can effectively reduce the noise level without adding additional physical burden.

[0003] However, in the process of achieving efficient active noise reduction, researchers face many challenges. First of all, how to accurately collect and separate multi-channel noise signals is one of the primary difficulties. The noise sources in the environment are often complex and changeable, which makes it extremely difficult to accurately extract the target noise source from many sound sources. Secondly, even if the noise signal can be successfully separated, it is also a technical difficulty to effectively locate and focus its direction. Because different noise sources may have similar frequency characteristics, and their directional information may also change at any time, this places high demands on the acoustic positioning algorithm.

[0004] In addition, in order to generate an effective reverse noise waveform to achieve phase reversal compensation, the directional noise signal must be accurately analyzed in the time and frequency domains. However, the existing technology still has deficiencies in balancing real-time processing and accuracy, resulting in sometimes unsatisfactory noise reduction effects. At the same time, the noise characteristics in different application scenarios vary greatly, which also requires the active noise reduction system to have a high degree of adaptability and flexibility. Therefore, how to improve the overall performance of the active noise reduction method based on the SOC chip has become an important direction of current research. Summary of the invention

[0005] The main purpose of the present invention is to provide an active noise reduction method and device based on a SOC chip, which solves the technical problem of how to accurately collect and separate multi-channel noise signals and reduce noise.

[0006] To achieve the above object, the present invention provides an active noise reduction method based on a SOC chip, comprising the following steps: The ambient noise is collected by using a preset microphone array to obtain a multi-channel noise signal; Performing blind source separation on the multi-channel noise signal to obtain separated noise sub-signals; Acquire a target noise source and direction information corresponding to the target noise source, and use an acoustic positioning algorithm in a target SOC chip to focus the target noise source based on the separated noise sub-signals and direction information to obtain a directional noise signal; Performing a time-frequency domain joint analysis on the directional noise signal through a target SOC chip to obtain time-frequency characteristic information of the directional noise signal; Based on the time-frequency characteristic information, reverse noise generation is performed on the directional noise signal to obtain a reverse noise waveform; Phase reversal compensation is performed on the directional noise signal based on the reverse noise waveform to obtain a noise-reduced target signal.

[0007] Furthermore, the method of collecting environmental noise using a preset microphone array to obtain a multi-channel noise signal includes: Using a preset microphone array to perform multi-dimensional sampling of ambient noise to obtain a time-domain discrete sampling sequence, and performing multi-channel synchronous quantization encoding on the time-domain discrete sampling sequence to obtain an initial digital signal; Channel cross-analysis is performed on the initial digital signal to obtain multi-channel noise spatial correlation characteristics, and noise field reconstruction processing is performed on the multi-channel noise spatial correlation characteristics to obtain a multi-channel noise signal.

[0008] Furthermore, performing blind source separation on the multi-channel noise signal to obtain separated noise sub-signals includes: Performing wavelet packet decomposition processing on the multi-channel noise signal to obtain a multi-level frequency band coefficient matrix, and performing energy density analysis on the multi-channel noise signal based on the multi-level frequency band coefficient matrix to obtain a frequency band energy distribution spectrum of the noise signal; Performing joint probability density estimation on the multi-channel noise signal based on the frequency band energy distribution spectrum to obtain a noise source statistical characteristic vector, and performing high-order cumulant analysis on the noise source statistical characteristic vector to obtain a signal source independence evaluation index; The independent components of the multi-channel noise signal are extracted and separated by using the information source independence evaluation index to obtain the separated noise sub-signals.

[0009] Furthermore, the target noise source and the direction information corresponding to the target noise source are obtained, and the target noise source is directionally focused based on the separated noise sub-signals and the direction information through the acoustic positioning algorithm in the target SOC chip to obtain a directional noise signal, including: Extracting acoustic features from the separated noise sub-signals to obtain an acoustic feature parameter set; wherein the acoustic feature parameter set includes a sound pressure level, a sound intensity vector, and a sound field gradient; Based on the acoustic characteristic parameter set, the noise sub-signal is spatially reconstructed to obtain a three-dimensional sound field distribution matrix, and the target noise source is subjected to sound source localization analysis based on the three-dimensional sound field distribution matrix and the direction information through an acoustic localization algorithm in a target SOC chip to obtain spatial coordinate information of the target noise source; wherein the spatial coordinate information includes azimuth, pitch angle and radial distance; The acoustic beam pattern is optimized by using a preset beam forming technology on the spatial coordinate information of the target noise source to obtain an optimal beam response function, and the noise sub-signal is spatially filtered based on the optimal beam response function to obtain a beam enhancement signal; Predicting the sound source trajectory of the target noise source based on the beam enhancement signal to obtain a sound source motion feature vector, and adaptively tracking the sound source motion feature vector to obtain a dynamic focusing parameter; The target noise source is spatially enhanced by using the dynamic focusing parameter to obtain a directional noise signal; wherein the directional noise signal includes amplitude spectrum and phase spectrum information of the target sound source.

[0010] Furthermore, the performing a time-frequency domain joint analysis on the directional noise signal to obtain time-frequency feature information of the directional noise signal includes: Performing a multiscale wavelet packet transform on the directional noise signal to obtain a multiscale wavelet coefficient matrix, and performing instantaneous frequency estimation on the directional noise signal based on the multiscale wavelet coefficient matrix to obtain an instantaneous frequency spectrum; wherein the instantaneous frequency spectrum includes frequency variation information of the directional noise signal in different frequency bands; Performing Hilbert-Huang transform on the instantaneous frequency spectrum to obtain an intrinsic mode function set, and performing edge detection on the instantaneous frequency spectrum based on the intrinsic mode function set to obtain an instantaneous frequency edge spectrum; Performing an S transform on the instantaneous frequency edge spectrum to obtain a time-frequency spectrum diagram, and tracking the energy center of gravity of the instantaneous frequency edge spectrum based on the time-frequency spectrum diagram to obtain an energy center of gravity trajectory; wherein the energy center of gravity trajectory includes energy distribution change information of the directional noise signal in the time-frequency domain; Performing Wigner-Ville distribution transformation on the energy center of gravity trajectory to obtain a Wigner-Ville distribution matrix, and performing ridge extraction on the Wigner-Ville distribution matrix to obtain a ridge frequency set; The Wigner-Ville distribution is subjected to time-frequency peak detection based on the ridge frequency set to obtain a time-frequency peak spectrum, and the time-frequency peak spectrum is parameterized and modeled to obtain time-frequency characteristic information of the directional noise signal.

[0011] Furthermore, the generating reverse noise on the directional noise signal based on the time-frequency characteristic information to obtain a reverse noise waveform includes: Performing complex cepstrum analysis on the directional noise signal based on the amplitude spectrum and the phase spectrum in the time-frequency feature information to obtain a complex cepstrum coefficient matrix, and reconstructing the phase spectrum of the directional noise signal based on the complex cepstrum coefficient matrix to obtain reconstructed phase spectrum information; Based on the reconstructed phase spectrum information, a minimum phase sequence is generated for the directional noise signal to obtain a minimum phase sequence, and based on the minimum phase sequence, a logarithmic spectrum correction is performed on the directional noise signal to obtain a corrected logarithmic spectrum; Performing spectrum inversion on the directional noise signal based on the corrected logarithmic spectrum to obtain an inverted spectrum, and performing complex frequency domain inverse transformation on the directional noise signal based on the inverted spectrum to obtain an initial inverse noise waveform; Performing time-frequency synchronization error analysis on the directional noise signal based on the initial reverse noise waveform to obtain a time-frequency synchronization error matrix, and performing phase error compensation on the initial reverse noise waveform based on the time-frequency synchronization error matrix to obtain a compensated reverse noise waveform; Performing sound pressure level matching on the directional noise signal based on the compensated reverse noise waveform to obtain matched sound pressure level parameters, and performing dynamic range compression on the compensated reverse noise waveform based on the matched sound pressure level parameters to obtain a reverse noise waveform with dynamic range compression; The directional noise signal is subjected to time-frequency domain energy equalization based on the reverse noise waveform after dynamic range compression to obtain an energy-equalized reverse noise waveform, and the directional noise signal is subjected to time-frequency domain phase inversion based on the energy-equalized reverse noise waveform to obtain an reverse noise waveform.

[0012] Furthermore, performing phase reversal compensation on the directional noise signal based on the reverse noise waveform to obtain a noise-reduced target signal includes: Performing time-frequency domain phase alignment on the directional noise signal based on the reverse noise waveform to obtain a phase-aligned noise signal, and performing phase inversion processing on the phase-aligned noise signal to obtain a phase-inverted noise signal, wherein the phase-inverted noise signal includes a noise component with a phase opposite to that of the directional noise signal; Based on the phase-inverted noise signal, the directional noise signal is superimposed in the time-frequency domain to obtain a superimposed mixed signal, and the superimposed mixed signal is subjected to time-frequency domain energy cancellation analysis to obtain a residual signal after energy cancellation; Adaptively suppress the residual noise of the superimposed mixed signal based on the residual signal after the energy cancellation to obtain an initial target signal after noise reduction.

[0013] The present invention also provides an active noise reduction device based on a SOC chip, comprising: A collection module is used to collect environmental noise using a preset microphone array to obtain a multi-channel noise signal; A separation module, used for performing blind source separation on the multi-channel noise signal to obtain separated noise sub-signals; A positioning module, used to obtain the target noise source and the direction information corresponding to the target noise source, and to focus the direction of the target noise source based on the separated noise sub-signals and the direction information through the acoustic positioning algorithm in the target SOC chip to obtain a directional noise signal; An analysis module, used to perform a time-frequency domain joint analysis on the directional noise signal through a target SOC chip to obtain time-frequency characteristic information of the directional noise signal; wherein the time-frequency characteristic information includes an amplitude spectrum and a phase spectrum; A generating module, configured to generate reverse noise on the directional noise signal based on the time-frequency characteristic information to obtain a reverse noise waveform; The compensation module is used to perform phase reversal compensation on the directional noise signal based on the reverse noise waveform to obtain a noise-reduced target signal.

[0014] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0015] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0016] The present invention provides an active noise reduction method based on a SOC chip, comprising the following steps: using a preset microphone array to collect environmental noise to obtain a multi-channel noise signal; performing blind source separation on the multi-channel noise signal to obtain a separated noise sub-signal; obtaining a target noise source and direction information corresponding to the target noise source, and using an acoustic positioning algorithm in a target SOC chip to perform direction focusing on the target noise source based on the separated noise sub-signal and the direction information to obtain a directional noise signal; performing time-frequency domain joint analysis on the directional noise signal through the target SOC chip to obtain time-frequency feature information of the directional noise signal; performing reverse noise generation on the directional noise signal based on the time-frequency feature information to obtain a reverse noise waveform; performing phase reversal compensation on the directional noise signal based on the reverse noise waveform to obtain a denoised target signal, thereby solving the technical problem of how to accurately collect and separate the multi-channel noise signal and reduce the noise, realizing combining the direction information of the target noise source, and realizing directional focusing on the target noise source through the acoustic positioning algorithm in the SOC chip, and this method can accurately locate and centrally process specific noise sources in a complex noise environment, greatly improving the noise reduction efficiency and effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the steps of an active noise reduction method based on a SOC chip in one embodiment of the present invention; Figure 2 is a structural block diagram of an active noise reduction device based on a SOC chip in one embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0018] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0020] like Figure 1 As shown, Figure 1 This is a schematic diagram of the steps of an active noise reduction method based on a SOC chip in one embodiment of the present invention; An embodiment of the present invention provides an active noise reduction method based on a SOC chip, comprising the following steps: Step S1, using a preset microphone array to collect environmental noise to obtain a multi-channel noise signal.

[0021] Specifically, using a preset microphone array to collect environmental noise and obtain a multi-channel noise signal is a key step in the active noise reduction method based on the SOC chip. Specifically, this step first requires deploying one or more pre-set microphone arrays in the target environment. These microphones are arranged in a specific geometric layout to ensure that the sound information in the environment can be captured from different angles and positions. In this way, sound signals from all directions can be collected to form a multi-channel noise signal set. The data captured by each microphone contains information from different sound sources, which includes both the target noise we want to eliminate and other background noise. For example, in an open office environment, employees are often disturbed by various noises, such as conversations between colleagues, the sound of printers, and the low-frequency noise of air conditioners. In order to improve this situation, an array consisting of multiple microphones can be installed in different corners of the office, so that sound signals from all directions can be collected at the same time. The signals received by these microphones will be transmitted to the SOC chip to prepare for the subsequent blind source separation processing. Due to the different positions of each microphone, the sound signals they receive will have slight time differences and intensity differences, which is the basis for achieving accurate noise separation. By analyzing these multi-channel noise signals, the system can more accurately identify which sounds are the target noises that need to be eliminated, thus laying a solid foundation for the next step of processing. Therefore, the effective execution of this initial step is crucial to the success of the entire active noise reduction process. It not only determines the quality of subsequent processing, but also directly affects the final noise reduction effect. In this process, every detail of the design, such as the selection and layout of microphones and data synchronization strategies, needs to be carefully considered to ensure that the collected multi-channel noise signals reflect the acoustic characteristics of the actual environment as realistically and comprehensively as possible. In this way, it can effectively support the further processing and optimization of noise signals in subsequent steps.

[0022] Step S2, performing blind source separation on the multi-channel noise signal to obtain separated noise sub-signals.

[0023] Specifically, the process of blind source separation of the multi-channel noise signal to obtain the separated noise sub-signal is one of the key steps in the active noise reduction method based on the SOC chip, and its purpose is to separate each independent noise source from the mixed noise signal captured by multiple microphones. In actual operation, the blind source separation technology relies on statistics and signal processing algorithms to analyze the correlation and difference between multi-channel noise signals, so as to identify and separate different noise sources. Since the sounds emitted by different sound sources have their own unique frequency characteristics and time characteristics, these characteristics can be captured in the multi-channel noise signal and deconstructed through complex mathematical models. Taking an open office environment as an example, assuming that there are multiple noise sources in this environment, such as employee conversations, printer operation, and air conditioning operation, the sound generated by each sound source will be recorded through the microphone array and form a multi-channel noise signal set. The blind source separation algorithm uses the time delay, amplitude change and other statistical characteristics between these signals to infer the number of original sound sources and their corresponding signal components. For example, the sound of conversation usually has a higher frequency and irregular time pattern, while the sound of the printer operation may show periodic pulse characteristics. By analyzing these features, the algorithm can separate these sounds from the mixed signal and generate separate noise sub-signals. This process requires powerful computing power and precise algorithm support, especially when facing complex and changing noise environments. With its integrated hardware architecture and efficient processing capabilities, the SOC chip is able to run these complex separation algorithms in real time. Once the individual noise sub-signals are successfully separated, the system can formulate more effective noise reduction strategies for each specific noise source, such as more precise focusing and suppression of noise in a specific direction or type. This not only improves the overall effect of noise reduction, but also provides high-quality input data for subsequent steps such as directional focusing and reverse noise generation, ensuring the effectiveness and accuracy of the entire active noise reduction process. Therefore, the successful implementation of the blind source separation step is crucial to achieving efficient active noise reduction.

[0024] Step S3, obtaining the target noise source and the direction information corresponding to the target noise source, and using the acoustic positioning algorithm in the target SOC chip to focus the target noise source based on the separated noise sub-signals and the direction information to obtain a directional noise signal.

[0025] Specifically, the target noise source and its corresponding direction information are obtained, and the target noise source is focused on the direction based on the separated noise sub-signal and direction information through the acoustic positioning algorithm in the target SOC chip to obtain the directional noise signal. This process is one of the core steps in the active noise reduction method. First, after completing the blind source separation, the system has obtained each independent noise sub-signal. The next task is to determine which noise sources are the target noises we need to process, and further identify the specific locations and directions of these noise sources. This step relies on the acoustic positioning algorithm, which uses the separated noise sub-signals and the spatial layout of the microphone array to calculate the direction information of each noise source. Taking the open office environment as an example, assuming that we have successfully separated different noise sub-signals such as conversation sounds, printer working sounds, and air conditioner running sounds, now we need to perform effective noise reduction processing for a specific noise source, such as the conversation sounds between employees. By analyzing the time difference and phase difference between the signals received by each microphone, the acoustic positioning algorithm in the SOC chip can accurately calculate the source direction of the conversation sound. For example, if the conversation sound mainly comes from the northeast corner of the office, the algorithm will be able to accurately identify this direction information and associate it with the corresponding noise sub-signal. Once the target noise source and its direction are determined, the system can focus on the noise signal in that direction, which is called "directional focusing". This means that the system will focus on processing all noise signals from this specific direction and ignore interference from other directions. The benefit of this is that the noise reduction effect can be significantly improved because the system's resources are dedicated to processing the most interfering noise sources. In this process, the powerful computing power of the SOC chip ensures that high-precision directional focusing operations can be completed in real time even in complex multi-noise environments. Ultimately, by focusing on the direction of the target noise source, the system can obtain a clearer and more concentrated directional noise signal, providing a high-quality data foundation for subsequent joint analysis in the time-frequency domain and reverse noise generation, thereby achieving more efficient active noise reduction. The effective execution of this series of steps not only improves the overall performance of noise reduction, but also creates a quieter and more comfortable working environment for users.

[0026] Step S4: performing a joint analysis of the directional noise signal in the time and frequency domains through the target SOC chip to obtain time and frequency feature information of the directional noise signal.

[0027] Specifically, the target SOC chip performs a joint analysis of the directional noise signal in the time and frequency domain to obtain the time and frequency characteristic information of the directional noise signal. This process is a key link in the active noise reduction method, which aims to deeply analyze the characteristics of the noise signal for subsequent processing. First, after completing the directional focusing and obtaining the directional noise signal, the next task is to perform a detailed time and frequency domain analysis on these signals. The joint analysis of the time and frequency domain combines the information of the time domain and the frequency domain, so that the system can fully understand the changing rules of the noise signal at different times and frequencies. Specifically, the SOC chip decomposes the directional noise signal into multiple time and frequency units, each of which represents a frequency component in a period of time, thereby revealing the temporal dynamic characteristics and frequency distribution of the noise signal. Taking the open office environment as an example, assume that we have successfully separated and focused the directional noise signal from the conversation of employees in the northeast corner. In order to effectively eliminate this noise, we need to understand its time and frequency characteristics in detail. The SOC chip can accurately extract the frequency composition of the noise signal in each time period by performing short-time Fourier transform (STFT) or other advanced time and frequency analysis techniques on the directional noise signal. For example, the conversation sound usually contains multiple frequency components, and these frequency components fluctuate over time. Through the joint analysis of the time and frequency domains, we can find that some high-frequency components may correspond to the vowel part of the speech, while the low-frequency components may be background noise or the consonant part of the speech. This detailed time-frequency feature analysis not only helps to identify the specific composition of the noise signal, but also provides the necessary data support for generating an effective reverse noise waveform. The powerful computing power of the SOC chip ensures that the joint analysis of the time and frequency domains can be completed efficiently even when processing complex and changeable noise signals in real time. Based on the obtained time-frequency feature information, the system can further optimize the reverse noise generation strategy to ensure that the generated reverse noise waveform can accurately match and offset the original noise signal. In this way, the entire active noise reduction process can be executed more accurately and effectively, significantly improving the noise reduction effect, and ultimately creating a quieter and more comfortable working environment for users. In addition, by continuously optimizing the joint analysis algorithm in the time and frequency domains, the system can also adapt to more diverse noise scenarios, further expanding its scope of application and effectiveness.

[0028] Step S5: generating reverse noise on the directional noise signal based on the time-frequency characteristic information to obtain a reverse noise waveform.

[0029] Specifically, the process of generating reverse noise for the directional noise signal based on the time-frequency feature information to obtain a reverse noise waveform is one of the key steps in the active noise reduction method, which aims to achieve effective noise cancellation by generating reverse noise with a phase opposite to the original noise signal. First, after completing the joint analysis of the time-frequency domain and obtaining the detailed time-frequency feature information of the directional noise signal, the system needs to use these feature data to generate an accurately matched reverse noise waveform. Specifically, the SOC chip calculates the corresponding reverse signal according to the frequency components and amplitude changes of the noise signal in each time-frequency unit, and its phase is completely opposite to the original noise signal, thereby achieving mutual cancellation of noise. Taking an open office environment as an example, assume that we have successfully separated, focused and analyzed the directional noise signal from the northeast corner of the employee's conversation. In order to effectively eliminate this noise, we need to generate a reverse noise waveform with a phase opposite to it. By deeply analyzing the time-frequency characteristics of the directional noise signal, the SOC chip can identify the specific time and intensity distribution of each frequency component in the conversation. For example, the conversation may contain a high-frequency vowel part and a low-frequency consonant part, and the amplitude and frequency of each part vary in different time periods. Based on these detailed time-frequency feature information, the SOC chip can generate reverse noise waveforms one by one, ensuring that each frequency component can be accurately matched and generate the corresponding reverse signal. In the process of generating the reverse noise waveform, the SOC chip uses a complex algorithm to adjust the phase and amplitude of the signal in each time-frequency unit, so that the generated reverse noise can be perfectly aligned with the original noise signal in time and frequency. In this way, when the original noise signal and the reverse noise waveform are superimposed, they will cancel each other out at the corresponding time points and frequencies, thereby significantly reducing the noise level. For example, in an open office environment, by generating and playing a reverse noise waveform with a phase opposite to that of the employee's conversation, the interference of the conversation to other employees can be effectively reduced, creating a quieter working environment. In addition, since the reverse noise generation relies on precise time-frequency feature information, it can adapt to different noise scenarios and adjust in real time to cope with changing noise conditions, further improving the stability and reliability of the noise reduction effect. Ultimately, the efficient execution of this series of steps not only improves the overall noise reduction performance, but also provides users with a better acoustic experience.

[0030] Step S6: performing phase reversal compensation on the directional noise signal based on the reverse noise waveform to obtain a noise-reduced target signal.

[0031] Specifically, the process of performing phase reversal compensation on the directional noise signal based on the reverse noise waveform to obtain the target signal after noise reduction is the last step in the active noise reduction method, which aims to achieve the best noise elimination effect through precise phase adjustment. After the generation of the reverse noise waveform is completed, the system needs to superimpose these reverse noise waveforms with the original directional noise signal to ensure that the two are completely aligned in time and frequency and have opposite phases. Specifically, the SOC chip will fine-tune the reverse noise waveform according to the time delay and phase information of the reverse noise waveform so that the reverse noise in each time-frequency unit can accurately offset the corresponding directional noise signal component. Taking the open office environment as an example, assume that we have successfully separated, focused and analyzed the directional noise signal from the employee conversation in the northeast corner, and generated the corresponding reverse noise waveform. In order to achieve effective noise reduction, we need to perform phase reversal compensation on the reverse noise waveform. This means that the SOC chip will calculate the phase difference between the reverse noise waveform and the original noise signal, and make necessary adjustments to ensure that the two can perfectly match in each time-frequency unit. For example, some high-frequency parts of the conversation may have a short period and a rapidly changing phase, while the low-frequency parts may exhibit a long period and relatively stable phase characteristics. Through precise phase reversal compensation, the SOC chip can ensure that the generated reverse noise waveform can be fully aligned with the original noise signal at every time point and frequency and cancel each other out. In this process, the powerful computing power and efficient algorithm support of the SOC chip are crucial. It can not only process complex noise signals in real time, but also dynamically adjust the compensation strategy according to environmental changes. For example, in an open office environment, when the volume or frequency of employees' conversations changes, the SOC chip can respond quickly, recalculate and adjust the phase and amplitude of the reverse noise waveform to ensure a continuous and effective noise reduction effect. Finally, after phase reversal compensation, the system outputs a target signal with significantly reduced noise interference. This not only improves the overall noise reduction performance, but also provides users with a clearer and quieter sound environment. Through this precise phase reversal compensation technology, the entire active noise reduction process can be efficiently executed, thereby significantly improving the user's auditory experience, especially in a noisy working environment, which can effectively reduce the interference of background noise and improve work efficiency and comfort.

[0032] In a specific embodiment, the method of collecting environmental noise using a preset microphone array to obtain a multi-channel noise signal includes: Using a preset microphone array to perform multi-dimensional sampling of ambient noise to obtain a time-domain discrete sampling sequence, and performing multi-channel synchronous quantization encoding on the time-domain discrete sampling sequence to obtain an initial digital signal; Channel cross-analysis is performed on the initial digital signal to obtain multi-channel noise spatial correlation characteristics, and noise field reconstruction processing is performed on the multi-channel noise spatial correlation characteristics to obtain a multi-channel noise signal.

[0033] Specifically, a preset microphone array is used to perform multi-dimensional sampling of the ambient noise to obtain a time-domain discrete sampling sequence, and the time-domain discrete sampling sequence is subjected to multi-channel synchronous quantization encoding to obtain an initial digital signal; then, the initial digital signal is subjected to channel cross-analysis to obtain multi-channel noise spatial correlation features, and the multi-channel noise spatial correlation features are subjected to noise field reconstruction processing to obtain a multi-channel noise signal. This series of steps is the basis for achieving efficient active noise reduction, and its purpose is to provide high-quality data support for subsequent noise separation and noise reduction by accurately capturing and processing ambient noise. First, in an open office environment, assume that we deploy an array of 8 microphones, which are distributed in different locations of the office according to a specific geometric layout. Each microphone collects the sound in the surrounding environment at a sampling frequency of 44,100 times per second, thereby forming a time-domain discrete sampling sequence. This means that in just one second, each microphone can generate 44,100 data points, and the entire array can generate 352,800 data points (8 microphones × 44,100 samples). These high-density sampling data can fully cover various noise components from low frequency to high frequency, ensuring the accuracy of subsequent processing. Next, in order to convert these analog signals into digital signals for further processing, the system will perform multi-channel synchronous quantization encoding on the time-domain discrete sampling sequence. In this process, the sound signal captured by each microphone will be converted into binary digital form for easy storage and transmission. For example, assuming that each sampling point is quantized using 16 bits, each microphone will generate 705,600 bits of data per second (44,100 samples × 16 bits), and the entire array will generate 5,644,800 bits (8 microphones × 705,600 bits) of data stream. This efficient digitization process not only ensures the fidelity of the signal, but also provides a solid foundation for subsequent complex calculations. Subsequently, by performing channel cross-analysis on the initial digital signal, the system can identify the spatial correlation of noise signals between each microphone. Specifically, in an open office environment, due to the differences in the location and propagation path of different sound sources, there are slight time differences and amplitude changes between the noise signals received by each microphone. For example, a microphone closer to the source of a conversation may record a higher-intensity sound signal, while a microphone farther away may record a lower-intensity signal. By analyzing these spatial correlation features, the system can build a detailed noise field model that includes the distribution of noise in space and its propagation characteristics. Suppose that after analysis, it is found that the phase difference between the conversation sounds of employees in the northeast corner between different microphones is about 5 milliseconds, which indicates the relative position relationship between the sound source and the microphone array, which helps with subsequent directional focusing processing. Finally, based on the above-mentioned multi-channel noise spatial correlation characteristics, the system performs noise field reconstruction processing and finally obtains a multi-channel noise signal.This step enables the system to more accurately describe the distribution and propagation of noise in three-dimensional space, thereby providing more accurate data input for subsequent processing steps such as blind source separation. For example, in an open office environment, through noise field reconstruction processing, the system can clearly distinguish which noise comes from a specific direction or area, and formulate targeted noise reduction strategies. In this way, even in the face of a complex multi-noise source environment, the system can effectively reduce background noise and significantly improve the quietness and comfort of the working environment. In this way, not only the overall noise reduction performance is improved, but also a more ideal acoustic experience is created for users. Throughout the process, the design of every detail is crucial, from the layout of the microphone array to the quantization encoding of the signal, to the extraction of spatially related features and the reconstruction of the noise field, which together constitute an efficient and accurate active noise reduction process.

[0034] In a specific embodiment, performing blind source separation on the multi-channel noise signal to obtain separated noise sub-signals includes: Performing wavelet packet decomposition processing on the multi-channel noise signal to obtain a multi-level frequency band coefficient matrix, and performing energy density analysis on the multi-channel noise signal based on the multi-level frequency band coefficient matrix to obtain a frequency band energy distribution spectrum of the noise signal; Performing joint probability density estimation on the multi-channel noise signal based on the frequency band energy distribution spectrum to obtain a noise source statistical characteristic vector, and performing high-order cumulant analysis on the noise source statistical characteristic vector to obtain a signal source independence evaluation index; The independent components of the multi-channel noise signal are extracted and separated by using the information source independence evaluation index to obtain the separated noise sub-signals.

[0035] Specifically, the process of blind source separation of the multi-channel noise signal to obtain the separated noise sub-signal includes multiple steps, the first of which is to analyze the multi-channel noise signal using wavelet packet decomposition. Specifically, in an open office environment, it is assumed that we have collected and digitized multiple noise signals from different directions through a microphone array. These noise signals include multiple noise sources such as employee conversations, printer working sounds, and air conditioner running sounds. In order to separate each independent noise source from these mixed signals, the system first needs to perform wavelet packet decomposition on the multi-channel noise signal. During the wavelet packet decomposition process, the noise signal of each channel will be decomposed into multiple levels of frequency band coefficient matrices. For example, assuming that we decompose the noise signal of each channel into 4 levels of frequency bands, each level contains 8 sub-bands, then there will be a total of 32 frequency band coefficient matrices (4 levels × 8 sub-bands). By performing energy density analysis on these frequency band coefficient matrices, the system can calculate the energy distribution of the noise signal in different frequency bands to form the frequency band energy distribution spectrum of the noise signal. For example, in a typical open office environment, conversation sounds are usually concentrated in the mid- and high-frequency bands, while the sounds of printers and air conditioners may be distributed in the low- and mid-frequency bands. Through energy density analysis, the system can clearly identify these different frequency band characteristics. Based on the frequency band energy distribution spectrum, the next step is to perform joint probability density estimation on the multi-channel noise signal to obtain the statistical feature vector of the noise source. This step uses the energy distribution information of the noise signal in different frequency bands to infer its statistical characteristics. For example, suppose the system finds that the energy distribution of certain frequency bands shows obvious non-Gaussian distribution characteristics, which indicates that these frequency bands are likely to correspond to specific noise sources. Through joint probability density estimation, the system can obtain a set of noise source statistical feature vectors, which describe the energy distribution patterns of each noise source in different frequency bands and their mutual relationships. Next, the system performs high-order cumulant analysis on the noise source statistical feature vectors to obtain the source independence evaluation index. High-order cumulant analysis is an advanced technique for evaluating signal independence. It quantifies the degree of independence between different noise sources by calculating the high-order statistics of the signal. For example, in an open office environment, suppose the system finds through high-order cumulants that the noise signal on a certain frequency band has a low correlation with the noise signals in other frequency bands, which means that the frequency band may correspond to an independent noise source. Through this analysis, the system is able to generate a set of source independence evaluation indicators that reflect the degree of independence of each noise source in the signal space. Finally, the system extracts and separates the independent components of the multi-channel noise signal through the source independence evaluation indicators, and finally obtains the separated noise sub-signals. In this process, the powerful computing power of the SOC chip ensures that the separation task can be completed efficiently even in the face of a complex multi-noise source environment.For example, in an open office environment, suppose the system successfully separates different noise sub-signals such as the conversation of employees in the northeast corner, the sound of printers, and the sound of air conditioners. In this way, the system can not only accurately identify each noise source, but also provide high-quality data input for subsequent directional focusing and noise reduction processing. The entire process, from wavelet packet decomposition of multi-channel noise signals to energy density analysis, to joint probability density estimation and high-order cumulant analysis, finally realizes independent component extraction and separation, forming a complete and efficient blind source separation process. Through this precise processing method, the system can effectively separate each independent noise source in a complex and changeable noise environment, significantly improving the overall active noise reduction effect and reliability. In this way, users can enjoy a better auditory experience in a quieter and more comfortable working environment, while improving work efficiency and quality of life.

[0036] In a specific embodiment, the step of acquiring the target noise source and the direction information corresponding to the target noise source, and performing direction focusing on the target noise source based on the separated noise sub-signals and the direction information through an acoustic positioning algorithm in a target SOC chip to obtain a directional noise signal includes: Extracting acoustic features from the separated noise sub-signals to obtain an acoustic feature parameter set; wherein the acoustic feature parameter set includes a sound pressure level, a sound intensity vector, and a sound field gradient; Based on the acoustic characteristic parameter set, the noise sub-signal is spatially reconstructed to obtain a three-dimensional sound field distribution matrix, and the target noise source is subjected to sound source localization analysis based on the three-dimensional sound field distribution matrix and the direction information through an acoustic localization algorithm in a target SOC chip to obtain spatial coordinate information of the target noise source; wherein the spatial coordinate information includes azimuth, pitch angle and radial distance; The acoustic beam pattern is optimized by using a preset beam forming technology on the spatial coordinate information of the target noise source to obtain an optimal beam response function, and the noise sub-signal is spatially filtered based on the optimal beam response function to obtain a beam enhancement signal; Predicting the sound source trajectory of the target noise source based on the beam enhancement signal to obtain a sound source motion feature vector, and adaptively tracking the sound source motion feature vector to obtain a dynamic focusing parameter; The target noise source is spatially enhanced by using the dynamic focusing parameter to obtain a directional noise signal; wherein the directional noise signal includes amplitude spectrum and phase spectrum information of the target sound source.

[0037] Specifically, the separated noise sub-signals are subjected to acoustic feature extraction to obtain an acoustic feature parameter set; wherein the acoustic feature parameter set includes sound pressure level, sound intensity vector and sound field gradient. This process is one of the key steps in the active noise reduction method based on the target SOC chip, and aims to obtain detailed acoustic feature information by in-depth analysis of the separated noise sub-signals. For example, in an open office environment, assume that we have successfully separated the noise sub-signal from the conversation of employees in the northeast corner. In order to further process these noise sub-signals, the system first extracts acoustic features from them. Specifically, the system calculates the sound pressure level (usually expressed in decibels dB) of each noise sub-signal, which reflects the intensity of the sound; the sound intensity vector describes the direction and speed of sound propagation; and the sound field gradient provides information about changes in the sound field. Through these detailed acoustic feature parameters, the system can fully understand the characteristics of the noise source. Based on the acoustic characteristic parameter set, the noise sub-signal is reconstructed into a spatial sound field to obtain a three-dimensional sound field distribution matrix, and the target noise source is analyzed for sound source positioning based on the three-dimensional sound field distribution matrix and the direction information through the acoustic positioning algorithm in the target SOC chip to obtain the spatial coordinate information of the target noise source; wherein the spatial coordinate information includes azimuth, pitch angle and radial distance. In actual operation, the system uses the above-mentioned acoustic characteristic parameter set to reconstruct the distribution of the noise source in three-dimensional space. For example, in an open office environment, it is assumed that the system has obtained the sound pressure level, sound intensity vector and sound field gradient data of multiple noise sub-signals from different directions. With these data, the system can construct a detailed three-dimensional sound field distribution matrix, which describes the position of each noise source in space and their relationship. Then, the system uses the built-in acoustic positioning algorithm to analyze these data to determine the specific location of the target noise source. For example, the system may find that a noise source is located in the northeast corner, with an azimuth of about 45 degrees, a pitch angle close to the horizontal plane, and a radial distance of 5 meters. This precise spatial coordinate information provides an important basis for subsequent directional focusing processing. The spatial coordinate information of the target noise source is optimized for the acoustic beam pattern through a preset beamforming technology to obtain an optimal beam response function, and the noise sub-signal is spatially filtered based on the optimal beam response function to obtain a beam enhancement signal. At this stage, the system adjusts the parameters of the beamformer according to the spatial coordinate information of the target noise source to optimize the directivity and sensitivity of the sound beam. For example, in an open office environment, assume that the system has determined the specific location of a noise source. Through beamforming technology, the system can generate an optimal beam response function so that the sound beam can be focused on the noise source, thereby improving the signal-to-noise ratio. For example, assume that the system adjusts the beam width to 10 degrees and points it to the noise source in the northeast corner.After optimization, the system can obtain a beam enhancement signal, which significantly enhances the sound component of the target noise source and suppresses the interference noise from other directions. Based on the beam enhancement signal, the target noise source is predicted to have a sound source motion feature vector, and the sound source motion feature vector is adaptively tracked to obtain a dynamic focusing parameter. In this process, the system not only pays attention to the current noise source position, but also predicts its future movement trend. For example, in an open office environment, suppose an employee walks around the office, causing the position of the sound source of his conversation to change constantly. The system can predict the motion trajectory of the sound source by analyzing the changes in the beam enhancement signal and generate the corresponding sound source motion feature vector. These feature vectors describe the motion parameters of the sound source, such as the speed and acceleration. Then, the system uses an adaptive direction tracking algorithm to update the dynamic focusing parameters in real time to ensure that the sound beam always accurately points to the moving noise source. Finally, the target noise source is spatially enhanced by the dynamic focusing parameter to obtain a directional noise signal; wherein the directional noise signal includes the amplitude spectrum and phase spectrum information of the target sound source. In this final step, the system further enhances the spatial direction of the target noise source based on the dynamic focusing parameters. For example, in an open office environment, assume that the system has predicted that an employee will move from the northeast corner to the southeast corner in the next few seconds. By adjusting the dynamic focusing parameters in real time, the system can ensure that the sound beam always follows the employee's movement path to achieve the best noise suppression effect. The directional noise signal contains not only the amplitude spectrum of the target sound source, but also the phase spectrum information, which provides a high-quality data basis for subsequent reverse noise generation and phase reversal compensation. The entire process, from acoustic feature extraction to spatial sound field reconstruction, to beamforming and adaptive direction tracking, constitutes an efficient and accurate active noise reduction process, which significantly improves the overall noise reduction effect and user experience. In this way, users can enjoy a better hearing experience in a quieter and more comfortable working environment, while improving work efficiency and quality of life.

[0038] In a specific embodiment, the performing a joint analysis of the directional noise signal in the time and frequency domains to obtain the time and frequency characteristic information of the directional noise signal includes: Performing a multiscale wavelet packet transform on the directional noise signal to obtain a multiscale wavelet coefficient matrix, and performing instantaneous frequency estimation on the directional noise signal based on the multiscale wavelet coefficient matrix to obtain an instantaneous frequency spectrum; wherein the instantaneous frequency spectrum includes frequency variation information of the directional noise signal in different frequency bands; Performing Hilbert-Huang transform on the instantaneous frequency spectrum to obtain an intrinsic mode function set, and performing edge detection on the instantaneous frequency spectrum based on the intrinsic mode function set to obtain an instantaneous frequency edge spectrum; Performing an S transform on the instantaneous frequency edge spectrum to obtain a time-frequency spectrum diagram, and tracking the energy center of gravity of the instantaneous frequency edge spectrum based on the time-frequency spectrum diagram to obtain an energy center of gravity trajectory; wherein the energy center of gravity trajectory includes energy distribution change information of the directional noise signal in the time-frequency domain; Performing Wigner-Ville distribution transformation on the energy center of gravity trajectory to obtain a Wigner-Ville distribution matrix, and performing ridge extraction on the Wigner-Ville distribution matrix to obtain a ridge frequency set; The Wigner-Ville distribution is subjected to time-frequency peak detection based on the ridge frequency set to obtain a time-frequency peak spectrum, and the time-frequency peak spectrum is parameterized and modeled to obtain time-frequency characteristic information of the directional noise signal.

[0039] Specifically, the process of performing a joint analysis of the directional noise signal in the time and frequency domain to obtain the time and frequency characteristic information of the directional noise signal includes multiple complex steps, aiming to extract its detailed time and frequency characteristics through in-depth analysis of the noise signal. First, in an open office environment, it is assumed that we have successfully separated and focused the directional noise signal from the conversation of employees in the northeast corner. In order to further process these directional noise signals, the system will perform a multi-scale wavelet packet transform on them. In the multi-scale wavelet packet transform process, the directional noise signal is decomposed into wavelet coefficient matrices of multiple scales. For example, suppose we decompose the signal into 4-level wavelet packets, each level contains 8 sub-bands, then there will be a total of 32 different frequency bands (4 levels × 8 sub-bands). Through this decomposition, the system can obtain the detailed frequency components on each frequency band, thereby forming a multi-scale wavelet coefficient matrix. Next, based on this matrix, the system will estimate the instantaneous frequency of the directional noise signal and obtain the instantaneous frequency spectrum. The instantaneous frequency spectrum not only contains the frequency change information of the directional noise signal in different frequency bands, but also reflects the temporal dynamic characteristics of the signal. For example, in an open office environment, conversations often have fast-changing frequency components, while printers and air conditioners may exhibit relatively stable frequency characteristics. Through instantaneous frequency estimation, the system can identify these different frequency change patterns. Next, the system performs a Hilbert-Huang transform on the instantaneous frequency spectrum to obtain a set of intrinsic mode functions. The Hilbert-Huang transform is a powerful tool for extracting intrinsic modal components from complex signals. Through this transform, the system can decompose the instantaneous frequency spectrum into multiple intrinsic mode functions, each representing an independent component of the signal. For example, in an open office environment, suppose the system finds that one intrinsic mode function corresponds to the high-frequency vowel part of the employee's conversation, while another function corresponds to the low-frequency consonant part. Based on these intrinsic mode functions, the system can further perform edge detection on the instantaneous frequency spectrum to obtain an instantaneous frequency edge spectrum. Edge detection helps to highlight the significant feature points in the signal, thereby providing a clear data basis for subsequent analysis. The system then performs an S transform on the instantaneous frequency edge spectrum to generate a time-frequency spectrum. The S transform combines the advantages of short-time Fourier transform and wavelet transform, and can provide high-resolution representation in the time-frequency domain. For example, in an open office environment, the time-frequency spectrum generated by the S transform can clearly show the frequency distribution of employee conversations in different time periods. Based on the time-frequency spectrum, the system will track the energy center of gravity of the instantaneous frequency edge spectrum to obtain the energy center of gravity trajectory. The energy center of gravity trajectory reflects the energy distribution change information of the directional noise signal in the time-frequency domain. For example, suppose the system finds that the energy of the conversation sound in a certain time period is concentrated in the mid-high frequency band, and gradually moves to the high frequency band over time, which indicates that the sound has obvious frequency drift characteristics.Next, the system will perform a Wigner-Ville distribution transformation on the energy center of gravity trajectory to obtain a Wigner-Ville distribution matrix. Wigner-Ville distribution is an efficient time-frequency analysis tool that can accurately describe the distribution of signals in the time-frequency domain. Through ridge extraction technology, the system can extract a series of ridge frequency sets from the Wigner-Ville distribution matrix. These ridge frequency sets represent the main frequency components of the signal and their changing trends over time. For example, in an open office environment, suppose the system finds that a certain ridge frequency set corresponds to the main frequency components of employee conversations, and these frequencies have changed slightly over time. Finally, based on the ridge frequency set, the system will perform time-frequency peak detection on the Wigner-Ville distribution to obtain a time-frequency peak spectrum. The time-frequency peak spectrum provides the main peak position and intensity information of the signal in the time-frequency domain. Through parametric modeling, the system can finally obtain the time-frequency feature information of the directional noise signal. For example, in an open office environment, suppose the system obtains the specific time-frequency features of employee conversations through parametric modeling, including its main frequency components, time and amplitude changes, and other information. These detailed time-frequency feature information provides high-quality data support for subsequent reverse noise generation and phase reversal compensation, thereby significantly improving the overall active noise reduction effect and reliability. The entire process, from multi-scale wavelet packet transform to instantaneous frequency estimation, to Hilbert-Huang transform, S transform, Wigner-Ville distribution transform and ridge extraction, constitutes an efficient and accurate time-frequency domain joint analysis process, ensuring that the system can achieve effective noise suppression in complex and changeable noise environments. In this way, users can enjoy a better auditory experience in a quieter and more comfortable working environment, while improving work efficiency and quality of life.

[0040] In a specific embodiment, the step of performing reverse noise generation on the directional noise signal based on the time-frequency feature information to obtain a reverse noise waveform includes: Performing complex cepstrum analysis on the directional noise signal based on the amplitude spectrum and the phase spectrum in the time-frequency feature information to obtain a complex cepstrum coefficient matrix, and reconstructing the phase spectrum of the directional noise signal based on the complex cepstrum coefficient matrix to obtain reconstructed phase spectrum information; Based on the reconstructed phase spectrum information, a minimum phase sequence is generated for the directional noise signal to obtain a minimum phase sequence, and based on the minimum phase sequence, a logarithmic spectrum correction is performed on the directional noise signal to obtain a corrected logarithmic spectrum; Performing spectrum inversion on the directional noise signal based on the corrected logarithmic spectrum to obtain an inverted spectrum, and performing complex frequency domain inverse transformation on the directional noise signal based on the inverted spectrum to obtain an initial inverse noise waveform; Performing time-frequency synchronization error analysis on the directional noise signal based on the initial reverse noise waveform to obtain a time-frequency synchronization error matrix, and performing phase error compensation on the initial reverse noise waveform based on the time-frequency synchronization error matrix to obtain a compensated reverse noise waveform; Performing sound pressure level matching on the directional noise signal based on the compensated reverse noise waveform to obtain matched sound pressure level parameters, and performing dynamic range compression on the compensated reverse noise waveform based on the matched sound pressure level parameters to obtain a reverse noise waveform with dynamic range compression; The directional noise signal is subjected to time-frequency domain energy equalization based on the reverse noise waveform after dynamic range compression to obtain an energy-equalized reverse noise waveform, and the directional noise signal is subjected to time-frequency domain phase inversion based on the energy-equalized reverse noise waveform to obtain an reverse noise waveform.

[0041] Specifically, the process of generating reverse noise for the directional noise signal based on the time-frequency feature information to obtain the reverse noise waveform includes multiple complex steps, which aims to generate a reverse noise waveform with a phase opposite to the original noise signal through in-depth analysis and processing of the directional noise signal, thereby achieving effective noise cancellation. First, in an open office environment, it is assumed that we have successfully separated and focused the directional noise signal from the conversation of employees in the northeast corner, and obtained its time-frequency feature information through detailed joint analysis in the time-frequency domain. In the first step, the system performs complex cepstrum analysis on the directional noise signal based on the amplitude spectrum and phase spectrum in the time-frequency feature information to obtain a complex cepstrum coefficient matrix, and reconstructs the phase spectrum of the directional noise signal based on the complex cepstrum coefficient matrix to obtain the reconstructed phase spectrum information. Complex cepstrum analysis is a powerful tool that can convert signals from the frequency domain to the complex cepstrum domain, thereby revealing the hidden structure inside the signal. For example, in an open office environment, it is assumed that the amplitude spectrum of the directional noise signal is mainly concentrated in the mid-high frequency band, while the phase spectrum exhibits complex fluctuation characteristics. Through complex cepstrum analysis, the system can convert these complex frequency and phase information into a complex cepstrum coefficient matrix, which contains all the spectral components of the signal and their interrelationships. Then, based on this complex cepstrum coefficient matrix, the system will reconstruct the phase spectrum of the directional noise signal to restore its original phase information. For example, suppose the system finds that the phase spectrum in a certain period of time shows obvious periodic fluctuations. Through phase spectrum reconstruction, the system can accurately restore these phase changes to ensure the accuracy of subsequent processing. Next, the system generates a minimum phase sequence for the directional noise signal based on the reconstructed phase spectrum information to obtain a minimum phase sequence, and performs logarithmic spectrum correction on the directional noise signal based on the minimum phase sequence to obtain a corrected logarithmic spectrum. The minimum phase sequence is a special signal representation with a minimum phase delay characteristic, which makes it very useful in noise processing. For example, in an open office environment, suppose the system generates a minimum phase sequence corresponding to the main frequency components of employee conversations. Through logarithmic spectrum correction, the system can further optimize the distribution of these frequency components to ensure consistency and stability throughout the entire spectrum. For example, suppose the system finds that some frequency components have significant energy deviations before correction. Through the corrected logarithmic spectrum, the system can effectively balance these energy differences and improve the quality of the overall spectrum. Then, the system performs spectrum inversion on the directional noise signal based on the corrected logarithmic spectrum to obtain an inverted spectrum, and performs a complex frequency domain inverse transform on the directional noise signal based on the inverted spectrum to obtain an initial inverted noise waveform. Spectral inversion is one of the key steps in generating an inverted noise waveform, which generates an inverted signal with a phase opposite to the original signal by inverting the spectrum of the signal.For example, in an open office environment, assume that the system obtains an inversion spectrum that is completely opposite to the employee's conversation through spectrum inversion. Then, through the complex frequency domain inverse transform, the system can convert this inversion spectrum back to the time domain to generate an initial reverse noise waveform. For example, assume that the initial reverse noise waveform generated by the system is completely aligned with the original conversation signal on the time axis, but its phase is completely opposite, which provides high-quality basic data for subsequent noise cancellation. Subsequently, the system performs time-frequency synchronization error analysis on the directional noise signal based on the initial reverse noise waveform to obtain a time-frequency synchronization error matrix, and performs phase error compensation on the initial reverse noise waveform based on the time-frequency synchronization error matrix to obtain a compensated reverse noise waveform. Since the noise signal in the actual environment may have a slight time delay and phase offset, it is necessary to perform time-frequency synchronization error analysis. For example, in an open office environment, assume that the system finds that there is a time delay of about 5 milliseconds between the initial reverse noise waveform and the original noise signal. Through time-frequency synchronization error analysis, the system can calculate a detailed time-frequency synchronization error matrix and perform phase error compensation on the initial reverse noise waveform based on the matrix. For example, suppose that the system reduces the time delay to less than 1 millisecond through compensation, which significantly improves the matching degree between the reverse noise waveform and the original noise signal. Then, the system matches the sound pressure level of the directional noise signal based on the compensated reverse noise waveform to obtain the matched sound pressure level parameters, and performs dynamic range compression on the compensated reverse noise waveform based on the matched sound pressure level parameters to obtain the reverse noise waveform after dynamic range compression. Sound pressure level matching is an important step to ensure that the reverse noise waveform is consistent with the original noise signal strength. For example, in an open office environment, suppose the system finds that the sound pressure level of the employee's conversation is about 60 decibels, while the sound pressure level of the initial reverse noise waveform is only 55 decibels. Through sound pressure level matching, the system can adjust the amplitude of the reverse noise waveform to keep it consistent with the sound pressure level of the original noise signal. Then, through dynamic range compression, the system can further optimize the dynamic range of the reverse noise waveform to ensure that it can work effectively under different environmental conditions. For example, suppose the system compresses the dynamic range from 80 decibels to 60 decibels, which not only improves the stability and reliability of the system, but also ensures the adaptability of the reverse noise waveform in various application scenarios. Finally, the system performs time-frequency domain energy equalization on the directional noise signal based on the reverse noise waveform after dynamic range compression to obtain an energy-equalized reverse noise waveform, and performs time-frequency domain phase inversion on the directional noise signal based on the energy-equalized reverse noise waveform to obtain an reverse noise waveform. Energy equalization is an important step to ensure that the reverse noise waveform is evenly distributed in the time-frequency domain. For example, in an open office environment, suppose the system finds that the energy distribution of the reverse noise waveform in certain frequency bands is uneven, resulting in poor noise reduction effect.Through energy equalization in the time-frequency domain, the system can optimize the energy distribution of the reverse noise waveform to ensure its consistency throughout the entire spectrum. For example, suppose the system adjusts the energy distribution from an unbalanced state in the high and low frequency bands to a uniform distribution, which significantly improves the noise reduction effect. Finally, through phase inversion in the time-frequency domain, the system generates a reverse noise waveform with a phase completely opposite to the original noise signal, achieving efficient noise cancellation. Throughout the process, from complex cepstrum analysis to phase spectrum reconstruction, to minimum phase sequence generation, logarithmic spectrum correction, spectrum inversion, time-frequency synchronization error analysis, sound pressure level matching, dynamic range compression, as well as energy equalization and phase inversion, a complete and efficient reverse noise generation process is formed. Through this precise processing method, the system can achieve effective noise suppression in complex and changeable noise environments, significantly improving the overall active noise reduction effect and reliability. Users can enjoy a better auditory experience in a quieter and more comfortable working environment, while improving work efficiency and quality of life. For example, in an open office environment, the reverse noise waveform generated by the above series of steps can effectively eliminate multiple noise sources such as employee conversations, printer operation, and air conditioner operation, creating a more peaceful working atmosphere.

[0042] In a specific embodiment, performing phase reversal compensation on the directional noise signal based on the reverse noise waveform to obtain a noise-reduced target signal includes: Performing time-frequency domain phase alignment on the directional noise signal based on the reverse noise waveform to obtain a phase-aligned noise signal, and performing phase inversion processing on the phase-aligned noise signal to obtain a phase-inverted noise signal, wherein the phase-inverted noise signal includes a noise component with a phase opposite to that of the directional noise signal; Based on the phase-inverted noise signal, the directional noise signal is superimposed in the time-frequency domain to obtain a superimposed mixed signal, and the superimposed mixed signal is subjected to time-frequency domain energy cancellation analysis to obtain a residual signal after energy cancellation; Based on the residual signal after the energy cancellation, the superimposed mixed signal is adaptively subjected to residual noise suppression to obtain an initial target signal after noise reduction.

[0043] Specifically, the process of performing phase reversal compensation on the directional noise signal based on the reverse noise waveform to obtain the target signal after noise reduction includes multiple key steps, which aims to achieve efficient noise cancellation by accurately processing the directional noise signal and the reverse noise waveform. First, in an open office environment, it is assumed that we have successfully generated a reverse noise waveform with a phase opposite to the employee's conversation voice, and are ready to apply it to the directional noise signal to achieve effective noise reduction. In the first step, the system performs time-frequency domain phase alignment on the directional noise signal based on the reverse noise waveform to obtain a phase-aligned noise signal, and performs phase reversal processing on the phase-aligned noise signal to obtain a phase-reversed noise signal, wherein the phase-reversed noise signal includes a noise component with a phase opposite to that of the directional noise signal. Phase alignment is a key step to ensure that the reverse noise waveform is completely aligned with the original directional noise signal in time and frequency. For example, in an open office environment, it is assumed that a certain frequency component in the directional noise signal has a significant peak at a certain point on the time axis, and the peak of the corresponding frequency component in the reverse noise waveform appears at a slightly delayed position. Through time-frequency domain phase alignment, the system can accurately adjust the time position of the reverse noise waveform to fully align it with the directional noise signal. For example, suppose the system finds that the time delay of the reverse noise waveform needs to be adjusted to 2 milliseconds to ensure its perfect alignment with the directional noise signal on the time axis. Then, by performing phase inversion processing on the phase-aligned noise signal, the system can obtain a noise component with a phase completely opposite to the directional noise signal. This phase inversion processing ensures that when the two signals are superimposed, they can cancel each other out at the corresponding time points and frequencies, thereby significantly reducing the noise level. Next, the system performs time-frequency domain superposition on the directional noise signal based on the phase-inverted noise signal to obtain a superimposed mixed signal, and performs time-frequency domain energy cancellation analysis on the superimposed mixed signal to obtain a residual signal after energy cancellation. In this process, the system superimposes the phase-inverted noise signal with the original directional noise signal. For example, in an open office environment, assume that the directional noise signal is mainly concentrated in the mid-high frequency band, while the reverse noise waveform contains the same frequency components with opposite phases. Through time-frequency domain superposition, the system can add the two signals point by point in each time-frequency unit to generate a superimposed mixed signal. For example, suppose the system finds that the superposition effect in a certain frequency band is the most significant and can effectively cancel out most of the noise components. However, since the noise signal in the actual environment may contain complex multi-band components, there may still be some residual noise in the superimposed mixed signal. Therefore, the system needs to perform time-frequency domain energy cancellation analysis on the superimposed mixed signal to identify and extract these residual noise components.For example, suppose the system finds that the residual noise intensity in certain frequency bands is high through energy cancellation analysis, which indicates that these frequency bands may be the focus areas for subsequent processing. Subsequently, the system performs adaptive residual noise suppression on the superimposed mixed signal based on the residual signal after energy cancellation to obtain the initial target signal after noise reduction. Adaptive residual noise suppression is a dynamically adjusted processing method that aims to further optimize the noise reduction effect. For example, in an open office environment, suppose the system finds that after the initial superposition processing, there are still some low-frequency background noises that cannot be completely eliminated. In order to deal with this situation, the system dynamically adjusts its noise suppression strategy based on the residual signal after energy cancellation. For example, suppose the system finds that the residual noise in a certain period of time is mainly concentrated in the low frequency band and has relatively stable characteristics. The system can further reduce the impact of these residual noises by enhancing the noise suppression strength of the low frequency band. Specifically, suppose the system increases the noise suppression coefficient of the low frequency band from 0.5 to 0.8 through an adaptive algorithm, which significantly improves the noise reduction effect of this frequency band. In addition, the system can also use advanced technologies such as machine learning to continuously optimize its noise suppression strategy based on real-time monitoring data to ensure the best noise reduction performance in various complex environments. Throughout the entire process, from time-frequency domain phase alignment to phase reversal processing, to time-frequency domain superposition, energy offset analysis, and adaptive residual noise suppression, a complete and efficient noise reduction process is formed. For example, in an open office environment, assume that the system successfully suppresses multiple noise sources such as employee conversations, printer operation, and air conditioner operation through the above series of steps. Specifically, assume that the total noise level initially detected by the system is about 65 dB, and after preliminary reverse noise generation and phase reversal compensation, the noise level is reduced to 55 dB. Then, through time-frequency domain superposition and energy offset analysis, the system further reduces the noise level to 45 dB. Finally, through adaptive residual noise suppression, the system finally stabilizes the noise level below 40 dB, significantly improving the overall noise reduction effect. Through this precise processing method, the system can not only achieve effective noise suppression in complex and changeable noise environments, but also dynamically adjust its processing strategy according to real-time environmental changes to ensure that the best noise reduction performance is always provided. Users can enjoy a better hearing experience in a quieter and more comfortable working environment, while improving work efficiency and quality of life. For example, in an open office environment, the reverse noise waveform generated by the above series of steps can effectively eliminate multiple noise sources such as employee conversations, printer operation, and air conditioning operation, creating a quieter working atmosphere. This not only reduces the interference of background noise, but also improves employee concentration and job satisfaction, further enhancing the overall competitiveness of the enterprise. In short, the effective execution of this series of steps enables the entire active noise reduction process to run efficiently, significantly improving the user's auditory experience and the quality of the working environment.

[0044] The above describes the active noise reduction method based on the SOC chip in the embodiment of the present invention. The following describes the active noise reduction device based on the SOC chip in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, an active noise reduction device based on a SOC chip includes: The acquisition module 21 is used to collect environmental noise using a preset microphone array to obtain a multi-channel noise signal; A separation module 22, configured to perform blind source separation on the multi-channel noise signal to obtain separated noise sub-signals; A positioning module 23 is used to obtain a target noise source and direction information corresponding to the target noise source, and to focus the target noise source based on the separated noise sub-signals and direction information through an acoustic positioning algorithm in a target SOC chip to obtain a directional noise signal; The analysis module 24 is used to perform a time-frequency domain joint analysis on the directional noise signal through the target SOC chip to obtain time-frequency characteristic information of the directional noise signal; wherein the time-frequency characteristic information includes an amplitude spectrum and a phase spectrum; A generating module 25, configured to generate reverse noise on the directional noise signal based on the time-frequency characteristic information to obtain a reverse noise waveform; The compensation module 26 is used to perform phase reversal compensation on the directional noise signal based on the reverse noise waveform to obtain a target signal after noise reduction.

[0045] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0046] Reference Figure 3 The present invention also provides a computer device in an embodiment, wherein the internal structure of the computer device can be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0047] Those skilled in the art will understand that Figure 3The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0048] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0049] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0050] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0051] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An active noise reduction method based on a SOC chip, characterized in that: The following steps are involved: The ambient noise is collected by using a preset microphone array to obtain a multi-channel noise signal; Performing blind source separation on the multi-channel noise signal to obtain separated noise sub-signals; Acquire a target noise source and direction information corresponding to the target noise source, and use an acoustic positioning algorithm in a target SOC chip to focus the target noise source based on the separated noise sub-signals and direction information to obtain a directional noise signal; Performing a time-frequency domain joint analysis on the directional noise signal through a target SOC chip to obtain time-frequency characteristic information of the directional noise signal; Based on the time-frequency characteristic information, reverse noise generation is performed on the directional noise signal to obtain a reverse noise waveform; Phase reversal compensation is performed on the directional noise signal based on the reverse noise waveform to obtain a noise-reduced target signal.

2. The active noise reduction method based on SOC chip according to claim 1, characterized in that: The method of collecting environmental noise using a preset microphone array to obtain a multi-channel noise signal includes: Using a preset microphone array to perform multi-dimensional sampling of ambient noise to obtain a time-domain discrete sampling sequence, and performing multi-channel synchronous quantization encoding on the time-domain discrete sampling sequence to obtain an initial digital signal; Channel cross-analysis is performed on the initial digital signal to obtain multi-channel noise spatial correlation characteristics, and noise field reconstruction processing is performed on the multi-channel noise spatial correlation characteristics to obtain a multi-channel noise signal.

3. The active noise reduction method based on SOC chip according to claim 1, characterized in that: The step of performing blind source separation on the multi-channel noise signal to obtain separated noise sub-signals includes: Performing wavelet packet decomposition processing on the multi-channel noise signal to obtain a multi-level frequency band coefficient matrix, and performing energy density analysis on the multi-channel noise signal based on the multi-level frequency band coefficient matrix to obtain a frequency band energy distribution spectrum of the noise signal; Performing joint probability density estimation on the multi-channel noise signal based on the frequency band energy distribution spectrum to obtain a noise source statistical characteristic vector, and performing high-order cumulant analysis on the noise source statistical characteristic vector to obtain a signal source independence evaluation index; The independent components of the multi-channel noise signal are extracted and separated by using the information source independence evaluation index to obtain the separated noise sub-signals.

4. The active noise reduction method based on SOC chip according to claim 1, characterized in that: The step of acquiring the target noise source and the direction information corresponding to the target noise source, and focusing the target noise source in a direction based on the separated noise sub-signals and the direction information through an acoustic positioning algorithm in a target SOC chip to obtain a directional noise signal includes: Extracting acoustic features from the separated noise sub-signals to obtain an acoustic feature parameter set; wherein the acoustic feature parameter set includes a sound pressure level, a sound intensity vector, and a sound field gradient; Based on the acoustic characteristic parameter set, the noise sub-signal is spatially reconstructed to obtain a three-dimensional sound field distribution matrix, and the target noise source is subjected to sound source localization analysis based on the three-dimensional sound field distribution matrix and the direction information through an acoustic localization algorithm in a target SOC chip to obtain spatial coordinate information of the target noise source; wherein the spatial coordinate information includes azimuth, pitch angle and radial distance; The acoustic beam pattern is optimized by using a preset beam forming technology on the spatial coordinate information of the target noise source to obtain an optimal beam response function, and the noise sub-signal is spatially filtered based on the optimal beam response function to obtain a beam enhancement signal; Predicting the sound source trajectory of the target noise source based on the beam enhancement signal to obtain a sound source motion feature vector, and adaptively tracking the sound source motion feature vector to obtain a dynamic focusing parameter; The target noise source is spatially enhanced by using the dynamic focusing parameter to obtain a directional noise signal; wherein the directional noise signal includes amplitude spectrum and phase spectrum information of the target sound source.

5. The active noise reduction method based on SOC chip according to claim 1, characterized in that: The step of performing a time-frequency domain joint analysis on the directional noise signal to obtain time-frequency feature information of the directional noise signal includes: Performing a multiscale wavelet packet transform on the directional noise signal to obtain a multiscale wavelet coefficient matrix, and performing instantaneous frequency estimation on the directional noise signal based on the multiscale wavelet coefficient matrix to obtain an instantaneous frequency spectrum; wherein the instantaneous frequency spectrum includes frequency variation information of the directional noise signal in different frequency bands; Performing Hilbert-Huang transform on the instantaneous frequency spectrum to obtain an intrinsic mode function set, and performing edge detection on the instantaneous frequency spectrum based on the intrinsic mode function set to obtain an instantaneous frequency edge spectrum; Performing an S transform on the instantaneous frequency edge spectrum to obtain a time-frequency spectrum diagram, and tracking the energy center of gravity of the instantaneous frequency edge spectrum based on the time-frequency spectrum diagram to obtain an energy center of gravity trajectory; wherein the energy center of gravity trajectory includes energy distribution change information of the directional noise signal in the time-frequency domain; Performing Wigner-Ville distribution transformation on the energy center of gravity trajectory to obtain a Wigner-Ville distribution matrix, and performing ridge extraction on the Wigner-Ville distribution matrix to obtain a ridge frequency set; The Wigner-Ville distribution is subjected to time-frequency peak detection based on the ridge frequency set to obtain a time-frequency peak spectrum, and the time-frequency peak spectrum is parameterized and modeled to obtain time-frequency characteristic information of the directional noise signal.

6. The active noise reduction method based on SOC chip according to claim 1, characterized in that: The generating reverse noise on the directional noise signal based on the time-frequency feature information to obtain a reverse noise waveform includes: Performing complex cepstrum analysis on the directional noise signal based on the amplitude spectrum and the phase spectrum in the time-frequency feature information to obtain a complex cepstrum coefficient matrix, and reconstructing the phase spectrum of the directional noise signal based on the complex cepstrum coefficient matrix to obtain reconstructed phase spectrum information; Based on the reconstructed phase spectrum information, a minimum phase sequence is generated for the directional noise signal to obtain a minimum phase sequence, and based on the minimum phase sequence, a logarithmic spectrum correction is performed on the directional noise signal to obtain a corrected logarithmic spectrum; Performing spectrum inversion on the directional noise signal based on the corrected logarithmic spectrum to obtain an inverted spectrum, and performing complex frequency domain inverse transformation on the directional noise signal based on the inverted spectrum to obtain an initial inverse noise waveform; Performing time-frequency synchronization error analysis on the directional noise signal based on the initial reverse noise waveform to obtain a time-frequency synchronization error matrix, and performing phase error compensation on the initial reverse noise waveform based on the time-frequency synchronization error matrix to obtain a compensated reverse noise waveform; Performing sound pressure level matching on the directional noise signal based on the compensated reverse noise waveform to obtain matched sound pressure level parameters, and performing dynamic range compression on the compensated reverse noise waveform based on the matched sound pressure level parameters to obtain a reverse noise waveform with dynamic range compression; The directional noise signal is subjected to time-frequency domain energy equalization based on the reverse noise waveform after dynamic range compression to obtain an energy-equalized reverse noise waveform, and the directional noise signal is subjected to time-frequency domain phase inversion based on the energy-equalized reverse noise waveform to obtain an reverse noise waveform.

7. The active noise reduction method based on SOC chip according to claim 1, characterized in that: The performing phase reversal compensation on the directional noise signal based on the reverse noise waveform to obtain a noise-reduced target signal includes: Performing time-frequency domain phase alignment on the directional noise signal based on the reverse noise waveform to obtain a phase-aligned noise signal, and performing phase inversion processing on the phase-aligned noise signal to obtain a phase-inverted noise signal, wherein the phase-inverted noise signal includes a noise component with a phase opposite to that of the directional noise signal; Based on the phase-inverted noise signal, the directional noise signal is superimposed in the time-frequency domain to obtain a superimposed mixed signal, and the superimposed mixed signal is subjected to time-frequency domain energy cancellation analysis to obtain a residual signal after energy cancellation; Based on the residual signal after the energy cancellation, the superimposed mixed signal is adaptively subjected to residual noise suppression to obtain an initial target signal after noise reduction.

8. An active noise reduction device based on a SOC chip, characterized in that: include: A collection module is used to collect environmental noise using a preset microphone array to obtain a multi-channel noise signal; A separation module, used for performing blind source separation on the multi-channel noise signal to obtain separated noise sub-signals; A positioning module, used to obtain the target noise source and the direction information corresponding to the target noise source, and to focus the direction of the target noise source based on the separated noise sub-signals and the direction information through the acoustic positioning algorithm in the target SOC chip to obtain a directional noise signal; An analysis module, used to perform a time-frequency domain joint analysis on the directional noise signal through a target SOC chip to obtain time-frequency characteristic information of the directional noise signal; wherein the time-frequency characteristic information includes an amplitude spectrum and a phase spectrum; A generating module, configured to generate reverse noise on the directional noise signal based on the time-frequency characteristic information to obtain a reverse noise waveform; The compensation module is used to perform phase reversal compensation on the directional noise signal based on the reverse noise waveform to obtain a noise-reduced target signal.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images