Sound console digital audio signal processing method and system, product and medium

Abstract

The invention discloses a sound console digital audio signal processing method and system, a product and a medium, and relates to the field of voice or sound processing, and the method comprises the steps: obtaining a plurality of paths of audio input signals, carrying out the time-frequency domain conversion, analyzing the frequency distribution rule, extracting the fundamental frequency and overtone structure features, and building a tone feature fingerprint database. Then obtaining a preset standard audio signal, extracting a multi-dimensional feature vector, and matching the multi-dimensional feature vector with the fingerprint database to obtain a feature parameter group; on the basis, a coordination index system including the overall sound coordination degree, the sound level definition and the spatial sound image balance degree is constructed. A coordination index of an input signal is calculated in real time and compared with a standard signal, and an adjustment value is obtained. And finally, in combination with fingerprint database matching features, calculating a target audio processing parameter combination containing balanced, dynamic and spatial parameters. By implementing the method, the sound quality of multiple paths of audio signals can be improved.

Description

Technical Field

[0001] This application relates to the field of voice or sound processing, and particularly to a method, system, product and medium for processing digital audio signals of a mixing console. Background Art

[0002] With the rapid development of modern music production and live performance technologies, the mixing console, as the core device for audio signal processing, plays an increasingly important role in fields such as music production, live performances, radio and television, etc. Digital mixing consoles can achieve precise adjustment and optimization of sounds by digitally processing multiple audio input signals to meet the professional audio processing requirements in different scenarios.

[0003] In related technologies, the audio signal processing of the mixing console adopts a fixed parameter preset method, and each channel is processed through a preset parameter template. This solution processes the input audio signals according to a fixed parameter combination through audio processing modules such as pre-configured equalizers, compressors, and reverbs, thereby achieving basic audio adjustment functions.

[0004] However, due to the differences in timbre characteristics between different audio signals, the preset fixed parameters are difficult to accurately analyze the overall coordination, layering, and spatial balance relationships among multiple audio signals, resulting in a poor overall sound quality. Summary of the Invention

[0005] This application provides a method, system, product and medium for processing digital audio signals of a mixing console, which is used to improve the sound quality of multiple audio signals.

[0006] First aspect, the present application provides a method for processing digital audio signals of a mixing console, which is applied to a digital audio signal processing system of a mixing console. The method includes: obtaining multiple channels of audio input signals, performing time-frequency domain conversion on the audio input signals to obtain time-frequency feature maps; analyzing the frequency distribution law of each audio input signal based on the time-frequency feature maps, extracting the fundamental frequency and overtone structure features of each audio input signal, and establishing a timbre feature fingerprint library according to the fundamental frequency and overtone structure features; obtaining a preset standard audio signal, extracting a multi-dimensional audio feature vector from the preset standard audio signal, and matching the multi-dimensional audio feature vector with the timbre feature fingerprint library to obtain an audio feature parameter group; constructing an audio feature coordination index system according to the audio feature parameter group, where the audio feature coordination index system includes overall sound coordination degree, voice part level clarity, and spatial sound image balance degree; calculating the audio feature coordination index of multiple channels of audio input signals in real time based on the audio feature coordination index system, comparing the audio feature coordination index with the coordination index of the preset standard audio signal to obtain a signal adjustment value; based on the signal adjustment value, calculating a target audio processing parameter combination in combination with the timbre feature fingerprint library, where the target audio processing parameter combination includes equalization parameters, dynamic parameters, and spatial parameters.

[0007] In the above embodiment, through time-frequency domain conversion and the establishment of a timbre feature fingerprint library, accurate extraction and recognition of the unique timbre features of each channel of audio signal are realized. Based on the feature matching between the timbre feature fingerprint library and the preset standard audio signal, an audio feature coordination index system including overall sound coordination degree, voice part level clarity, and spatial sound image balance degree is constructed, thus breaking through the limitation of fixed parameter preset. Dynamically comparing the coordination index calculated in real time with the standard signal, and calculating the target parameter combination in combination with the matching features of the timbre feature fingerprint library, realizing the adaptive processing of audio signals with different timbre features, and ensuring the optimization effect of multiple channels of audio signals in terms of overall coordination, layering, and spatial balance.

[0008] Combined with some embodiments of the first aspect, in some embodiments, the step of constructing an audio feature coordination index system according to the audio feature parameter group, where the index system includes overall sound coordination degree, voice part level clarity, and spatial sound image balance degree, specifically includes: calculating the energy values of different frequency bands in the audio feature parameter group, obtaining the energy ratio between adjacent frequency bands, and calculating the overall sound coordination degree index according to the energy ratio; extracting the voice part loudness values and frequency values in the audio feature parameter group, and calculating the voice part level clarity index corresponding to the overall sound coordination degree index; calculating the phase difference and amplitude difference of the channel signals according to the left and right channel signal values in the audio feature parameter group, and determining the sound image position distribution in combination with the voice part level clarity index to obtain the spatial sound image balance degree index.

[0009] In the above embodiments, the overall sound coordination index is obtained by calculating the energy ratios of different frequency bands, and this index reflects the balance of the spectral energy distribution. By combining the sound intensity of each voice part and the frequency values to analyze the differentiation degree of different voice parts, the clarity index of voice part levels is calculated, which reflects the sense of hierarchy between voice parts. Further, based on the phase difference and amplitude difference of the left and right channel signals, and in combination with the clarity of voice part levels, the sound image position distribution is determined, and finally a complete spatial sound image balance index is formed, enabling the audio signal processing to have a comprehensive coordination evaluation ability.

[0010] In combination with some embodiments of the first aspect, in some embodiments, the step of calculating the audio feature coordination index of the multi-channel audio input signals in real time based on the audio feature coordination index system and comparing the audio feature coordination index with the coordination index of the preset standard audio signal specifically includes: segmenting the multi-channel audio input signals at time intervals according to the overall sound coordination calculation rule, the clarity calculation rule of voice part levels, and the spatial sound image balance calculation rule in the audio feature coordination index system, and calculating the coordination index of each segment of the signal; extracting the coordination index of the corresponding time period from the preset standard audio signal, making a numerical comparison with the coordination index of the multi-channel audio input signals to obtain a numerical comparison result; and determining the signal adjustment value according to the numerical comparison result and the matching features of the timbre feature fingerprint library.

[0011] In the above embodiments, the multi-channel audio input signals are segmented at time intervals for processing, and the coordination index is calculated for each segment of the signal, establishing a dynamic evaluation mechanism for audio signal processing. By making a numerical comparison with the coordination index of the corresponding time period in the preset standard audio signal and combining the matching features of the timbre feature fingerprint library to determine the signal adjustment value, the precise adjustment of the audio signal processing parameters is realized, ensuring the continuity and accuracy of the processing results.

[0012] In combination with some embodiments of the first aspect, in some embodiments, after the step of calculating the target audio processing parameter combination in combination with the timbre feature fingerprint library based on the signal adjustment value, the method further includes: extracting the feature change rule of the preset standard audio signal and establishing a feature change record table; analyzing the adjustment direction of the target parameter combination according to the feature change record table to generate a parameter adjustment step size; and adjusting the target audio processing parameter combination according to the parameter adjustment step size.

[0013] In the above embodiments, extracting the feature change rule of the preset standard audio signal and establishing a feature change record table provides a dynamic reference benchmark for parameter adjustment. Analyzing the adjustment direction of the target parameter combination according to the feature change record table to generate a parameter adjustment step size realizes the precise control of parameter adjustment. Using the parameter adjustment step size to numerically adjust the target audio processing parameter combination makes the adjustment process of the audio processing parameters smoother and avoids the sound quality fluctuations caused by parameter mutations.

[0014] In combination with some embodiments of the first aspect, in some embodiments, after the steps of obtaining multi-channel audio input signals and performing time-frequency domain conversion on the audio input signals to obtain time-frequency feature maps, the method further includes: analyzing the timbre feature distribution in the time-frequency feature maps, identifying the signal features of different musical instruments and human voices according to the timbre feature distribution; separating the multi-channel audio input signals into independent audio tracks according to the signal features; extracting the fundamental frequency and overtone structure features of the independent audio tracks, and performing feature matching with the timbre feature fingerprint library; calculating the audio feature coordination index of the independent audio tracks to obtain the target parameter combination of each audio track.

[0015] In the above embodiments, the signal features of different musical instruments and human voices are distinguished based on the timbre feature distribution in the time-frequency feature maps, and the multi-channel audio input signals are separated into independent audio tracks. By extracting the fundamental frequency and overtone structure features of the independent audio tracks for timbre feature fingerprint matching and calculating the audio feature coordination index for each audio track separately, accurate identification and independent processing of different types of audio signals are achieved, improving the refinement degree of audio processing.

[0016] In combination with some embodiments of the first aspect, in some embodiments, after the step of adjusting the numerical value based on the signal and calculating the target audio processing parameter combination in combination with the timbre feature fingerprint library, the method further includes: extracting the frequency change feature and energy change feature in the time-frequency feature maps, and analyzing the change rules of the features; predicting the fundamental frequency change trend and overtone structure change trend in the next time period according to the change rules and the historical data of the audio feature parameter group; adjusting the numerical values of each index in the audio feature coordination index system based on the fundamental frequency change trend and overtone structure change trend; substituting the adjusted numerical values of each index into the calculation process of the target audio processing parameter combination to obtain a pre-adjusted parameter group including equalization parameters, dynamic parameters, and spatial parameters.

[0017] In the above embodiments, the change rules of the frequency and energy in the time-frequency feature maps are analyzed, and the fundamental frequency and overtone structure change trends in the next time period are predicted in combination with the historical data of the audio feature parameter group. Based on the prediction results, the numerical values of each index in the audio feature coordination index system are dynamically adjusted, and the adjusted numerical values of each index are applied to the parameter combination calculation, realizing the pre-adjustment of the audio processing parameters and enhancing the real-time performance and continuity of audio processing.

[0018] In combination with some embodiments of the first aspect, in some embodiments, after the step of calculating the target audio processing parameter combination based on the signal adjustment value and in combination with the timbre feature fingerprint library, the method also includes: obtaining an ambient noise signal, converting the ambient noise signal into a noise time-frequency feature graph, and extracting frequency distribution characteristics and energy distribution characteristics in the noise time-frequency feature graph; comparing the frequency distribution characteristics and energy distribution characteristics of the noise time-frequency feature graph with the time-frequency feature graph of the audio input signal to determine the noise frequency band and noise energy; constructing a noise suppression index based on the noise frequency band and noise energy, and adding the noise suppression index to the audio feature coordination index system; and adjusting the equalization parameters and dynamic parameters in the target audio processing parameter combination based on the value of the noise suppression index.

[0019] In the above embodiment, the environmental noise signal is obtained and converted into a noise time-frequency characteristic diagram, and its frequency distribution characteristics and energy distribution characteristics are compared with the time-frequency characteristic diagram of the audio input signal to accurately identify the noise frequency band and noise energy. Based on the recognition result, a noise suppression index is constructed and incorporated into the audio feature coordination index system, and the equalization parameters and dynamic parameters in the target audio processing parameter combination are adjusted in a targeted manner, thereby achieving accurate recognition and suppression of environmental noise and improving the anti-interference ability of audio signal processing in complex environments.

[0020] In a second aspect, an embodiment of the present application provides a mixing console digital audio signal processing system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the mixing console digital audio signal processing system to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0021] In a third aspect, an embodiment of the present application provides a computer program product comprising instructions. When the above-mentioned computer program product is run on a mixing console digital audio signal processing system, the above-mentioned mixing console digital audio signal processing system executes the method described in the first aspect and any possible implementation method of the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions. When the above instructions are executed on a mixing console digital audio signal processing system, the above mixing console digital audio signal processing system executes the method described in the first aspect and any possible implementation method of the first aspect.

[0023] Understandably, the mixer digital audio signal processing system provided in the second aspect above, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the method provided in the embodiments of the present application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method, and will not be elaborated here.

[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. Through time-frequency domain conversion and the establishment of a timbre feature fingerprint library in the present application, the accurate extraction and recognition of the unique timbre features of each audio signal are realized. Based on the feature matching between the timbre feature fingerprint library and the preset standard audio signal, an audio feature coordination index system including the overall sound coordination degree, the clarity of voice part levels, and the spatial sound image balance degree is constructed, thus breaking through the limitation of fixed parameter presetting. By dynamically comparing the coordination index calculated in real time with the standard signal and calculating the target parameter combination in combination with the matching features of the timbre feature fingerprint library, the adaptive processing of audio signals with different timbre features is realized, ensuring the optimization effect of multi-channel audio signals in terms of overall coordination, layering, and spatial balance.

[0025] 2. By calculating the energy ratio of different frequency bands in the present application, the overall sound coordination degree index is obtained, which reflects the balance of spectral energy distribution. Combining the voice part loudness and frequency values to analyze the differentiation degree of different voice parts, the clarity index of voice part levels is calculated, reflecting the layering between voice parts. Further, based on the phase difference and amplitude difference between the left and right channel signals, and combining with the clarity of voice part levels to determine the sound image position distribution, the complete spatial sound image balance degree index is finally formed, enabling the audio signal processing to have a comprehensive coordination evaluation ability.

[0026] 3. By segmenting the multi-channel audio input signals at time intervals and calculating the coordination index for each segment respectively in the present application, a dynamic evaluation mechanism for audio signal processing is established. By numerically comparing with the coordination index in the corresponding time period of the preset standard audio signal and determining the signal adjustment value in combination with the matching features of the timbre feature fingerprint library, the accurate adjustment of audio signal processing parameters is realized, ensuring the continuity and accuracy of the processing results. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a flowchart of a method for processing digital audio signals of a mixer in an embodiment of the present application; Figure 2 is another flowchart of a method for processing digital audio signals of a mixer in an embodiment of the present application; Figure 3 is a schematic structural diagram of an entity device of a mixer digital audio signal processing system in an embodiment of the present application. Detailed implementation manners

[0028] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. As used in the specification of the present application, the singular forms "a", "an", "above-mentioned", "the", and "this" are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations including one or more of the listed items.

[0029] Hereinafter, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as implying or suggesting relative importance or implicitly indicating the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0030] For ease of understanding, the application scenarios of the embodiments of the present application are introduced below.

[0031] During the live broadcast of large stadiums, there are various complex audio processing requirements. The venue party conducted a rehearsal recording before the event and obtained a standard audio signal in an ideal environment, including clear commentary sounds (-20 dBFS, 200 Hz - 8 kHz), moderate audience background sounds (-35 dBFS, 300 Hz - 3 kHz), and live music (-25 dBFS, 50 Hz - 16 kHz). The frequency response curve of this standard audio is smooth, the dynamic range is 90 dB, the signal-to-noise ratio is greater than 60 dB, and the spatial sound field distribution is uniform. However, in actual live broadcasts, due to the complex and variable environmental noise, ranging from -20 dBFS to -80 dBFS and with a wide frequency band distribution (20 Hz - 20 kHz), traditional fixed-parameter audio processing solutions are difficult to adjust the real-time audio to a level close to the standard audio. In addition, there is spectral overlap between different sound sources, such as the overlap of the audience's noise and the commentary sound in the mid-frequency band, which increases the difficulty of separating and enhancing the audio signal. At the same time, the on-site audio system also needs to handle the problem of blurred sound caused by the venue reverberation (RT60 reaches 2.5 seconds), as well as the problem of uneven frequency response caused by the spatial acoustic characteristics.

[0032] A certain studio uses a traditional audio processing system for live program production. Before the program starts, the production team collects a standard audio in a noise-free environment, which includes the host's standard voice (-18dBFS, 200Hz - 8kHz) and background music (-30dBFS, 50Hz - 16kHz). This audio has excellent frequency response characteristics and a dynamic range of 90dB. The system uses a fixed 31-band graphic equalizer for frequency response correction, adopts standard compressor settings (compression ratio of 2:1, threshold of -18dB) for dynamic range control, and uses preset reverb parameters (room size 80%, pre-delay 30ms) for spatial effect processing. When the program reaches the halfway point, due to the increase in low-frequency noise (100 - 300Hz, -45dBFS) generated by the studio air-conditioning system, the traditional system cannot optimize the parameters according to the standard audio characteristics, resulting in the low-frequency noise masking the host's voice. At the same time, due to the fixed compressor settings being unable to adaptively adjust according to the noise level, a pumping effect of the sound is caused. In addition, the preset reverb parameters of the system cannot be adjusted according to the actual acoustic environment changes, and instead increase the sense of turbidity of the sound in a noisy environment. The system also has a processing delay (about 50ms), resulting in a listening perception deviation in on-site monitoring. Finally, the audio quality of the program is significantly lower than the standard audio, and the clarity index drops from 0.85 to 0.65, and the dynamic range is compressed from 90dB to 60dB, seriously affecting the program effect.

[0033] A certain performance venue uses the audio processing system of this application for on-site sound reinforcement. Before the performance, the system obtained a standard audio in an ideal environment, including the lead singer's vocals (-15 dBFS, 100 Hz - 10 kHz), accompaniment music (-20 dBFS, 20 Hz - 20 kHz), and ambient reverberation (-40 dBFS). The system analyzed the characteristics of the standard audio and established an audio feature coordination index system. At the performance site, the system first detected three types of main noises through real-time analysis: the background noise in the audience area (500 - 2 kHz, -50 dBFS), the operating noise of stage equipment (above 4 kHz, -60 dBFS), and the low-frequency noise of the air-conditioning system (100 - 200 Hz, -45 dBFS). The system compared these noise characteristics with the characteristics of the standard audio, constructed noise suppression indexes, and calculated the spectral masking value and energy suppression value in 24 critical frequency bands. Based on these indexes, the system automatically adjusted the settings of a 31-band parametric equalizer: increased the Q value to 8.65 and set a -6 dB attenuation in the low-frequency band; used adaptive notch filtering with a Q value of 12.5 in the mid-frequency band; used a high-shelf filter with a slope of -12 dB / octave in the high-frequency band. The dynamic processor parameters were also optimized according to the dynamic characteristics of the standard audio: the compression ratio in the low-frequency band was increased to 3:1 and the threshold was reduced to -24 dB; 4:1 compression was used in the mid-frequency band with a fast attack time of 1 ms; an expander was used for processing in the high-frequency band. Through these adaptive adjustments, the system successfully adjusted the characteristics of the on-site audio to be close to the level of the standard audio, the signal-to-noise ratio was increased by 15 dB, the sound clarity index was maintained above 0.85, and the dynamic range was kept at 85 dB, effectively solving the sound quality problem in a complex acoustic environment.

[0034] For ease of understanding, the method provided in this embodiment will be described in terms of its process in combination with the above scenario. Please refer to Figure 1 , which is a schematic flowchart of a process of the digital audio signal processing method of the mixing console in the embodiment of this application.

[0035] S101. Obtain multiple audio input signals, and perform time-frequency domain conversion on the audio input signals to obtain a time-frequency feature map.

[0036] Among them, the multiple audio input signals represent multiple independent audio signal channels from different audio sources, including but not limited to instrument signals, vocal signals, etc.; the time-frequency domain conversion refers to converting a time-domain audio signal into a time-frequency representation form that simultaneously contains time and frequency information; the time-frequency feature map represents a two-dimensional image representation of the frequency distribution characteristics of the audio signal at different time points.

[0037] This step starts to execute after the system is started and an audio input device is detected. Specifically, the system collects audio signals of multiple channels through the audio input interface, performs analog-to-digital conversion on the audio signals of each channel to obtain digital audio signals. Then, the digital audio signals are framed, and time-frequency analysis methods such as short-time Fourier transform (STFT) are applied to each frame of the signal to obtain the distribution characteristics of the signal in the time-frequency plane, forming a time-frequency feature map.

[0038] In some embodiments, time-frequency domain signal processing can be implemented in the following manner: Optionally, first preprocess the input signal, including signal normalization, noise reduction, and framing, then window each frame of the signal and perform FFT transformation, and finally arrange the transformation results in chronological order to form a time-frequency diagram; Optionally, perform multi-resolution analysis on the signal using wavelet transform, and construct a time-frequency feature map through wavelet coefficients of different scales, which can better represent the local characteristics of the signal. It can be understood that other time-frequency analysis methods can also be used to implement the time-frequency representation of audio signals, which is not limited here.

[0039] After this step, the following steps are further included: Analyze the distribution of timbre features in the time-frequency feature map, and identify the signal features of different musical instruments and human voices according to the timbre feature distribution.

[0040] In this step, the time-frequency feature map refers to the time-frequency distribution map of the audio signal obtained through the short-time Fourier transform; the timbre feature distribution includes spectral envelope, transient features, and harmonic structure; the signal features refer to the typical acoustic feature parameters of different sound sources. The short-time Fourier transform uses a 2048-point FFT, a frame length of 25 ms, a frame shift of 12.5 ms, a Hanning window, and an analysis frequency range of 20 Hz - 20 kHz. The system performs a time-frequency analysis process on the input mixed audio signal. First, it generates a time-frequency feature map through the short-time Fourier transform to obtain the time evolution features of the spectrum. In the frequency domain, the system divides the frequency range of 20 Hz - 20 kHz into 24 critical bands and calculates the energy distribution of each band. For the human voice signal, the system identifies its characteristic formant distribution (F1 ≈ 500 Hz, F2 ≈ 1500 Hz, F3 ≈ 2500 Hz) and the typical pitch range (80 - 1000 Hz). For the piano signal, the system identifies its wide spectrum feature (27.5 Hz - 4186 Hz) and the typical harmonic structure (integer multiples of the fundamental frequency). For the guitar signal, the system identifies the mid-frequency energy concentration feature (80 - 1200 Hz) and the unique plucking transient feature (attack time < 5 ms). In the time domain, the system analyzes the envelope features of various signals: the gradual change feature of the human voice (attack time 20 - 50 ms), the rapid decay feature of the piano (decay time 500 - 1000 ms), and the medium decay feature of the guitar (decay time 200 - 500 ms). The system further quantifies the timbre features of different sound sources by calculating feature parameters such as Mel-frequency cepstral coefficients (MFCC) and spectral centroid.

[0041] Separate the multi-channel audio input signals into independent tracks according to the signal features.

[0042] In this step, the multi-channel audio input signals refer to the mixed multi-source audio signals; the independent tracks refer to the separated single-source audio signals; the signal separation uses a deep learning-based source separation algorithm. The system performs the audio signal separation process. First, it preprocesses the input mixed audio, including signal normalization and frequency domain transformation. The system uses a convolutional neural network model with a U-Net structure for signal separation. The input layer receives the time-frequency map of the mixed signal (frame length 25 ms, frame shift 12.5 ms), and extracts features and reconstructs the spectrograms of each sound source through a 5-layer encoder-decoder structure. In the time-frequency masking layer, the system generates a soft masking matrix for each target sound source, and the masking value range is 0 - 1. The masked spectrum is optimized through a Wiener filter to reduce artifacts during the separation process. Finally, the time-domain signal is reconstructed through the inverse short-time Fourier transform to obtain the independent tracks of the human voice, piano, and guitar. The system maintains the phase consistency of each track through a phase reconstruction algorithm and uses a residual compensation technique to improve the separation quality.

[0043] Extract the fundamental frequency and overtone structure features of the independent audio track, and perform feature matching with the timbre feature fingerprint library.

[0044] In this step, the fundamental frequency refers to the lowest frequency component of the audio signal; the overtone structure refers to the energy distribution of the integer multiple frequency components of the fundamental frequency; the timbre feature fingerprint library contains standard feature templates of various musical instruments and human voices. The system performs the feature extraction and matching process on each separated audio track. For each audio track, the system first extracts the fundamental frequency through the autocorrelation method, calculates the spectrum using a 2048-point FFT, and identifies the peak corresponding to the fundamental frequency. In the human voice track, the system extracts the pitch contour (fundamental frequency range 80 - 1000 Hz) and formant structure (positions and bandwidths of F1, F2, F3). In the piano track, the system analyzes the fundamental frequencies of 88 keys (27.5 - 4186 Hz) and the overtone attenuation characteristics (energy ratios of the first 8 overtones). In the guitar track, the system extracts the fundamental frequencies of 6 strings (82 - 698 Hz) and the overtone structure. The system matches the extracted feature vectors with the templates in the timbre feature fingerprint library, uses the Euclidean distance to measure the feature similarity, and selects the feature template with the highest similarity as the reference.

[0045] Calculate the audio feature coordination index of the independent audio track to obtain the target parameter combination of each audio track.

[0046] In this step, the audio feature coordination index includes spectral balance, dynamic consistency, and spatial correlation; the target parameter combination includes the parameter settings of the equalizer, dynamic processor, and spatial processor. The system performs the feature coordination analysis and parameter calculation process. For each audio track, the system first calculates the spectral balance index: by performing variance analysis on the energy distribution of 24 critical bands, calculate the smoothness of the frequency response curve. The spectral balance requirement for the human voice track in the range of 200 - 8000 Hz is greater than 0.85, for the piano track in the full frequency range is greater than 0.80, and for the guitar track in the range of 80 - 5000 Hz is greater than 0.75. The dynamic consistency index is obtained by calculating the change rate of the short-time energy envelope, including the attack time, release time, and dynamic range parameters. The spatial correlation index is obtained by analyzing the correlation coefficient and sound image localization characteristics of the stereo signal. The system calculates the processing parameters of each audio track based on these indexes: the center frequency, Q value, and gain settings of the 31-band parametric equalizer, the compression ratio, threshold, attack time, and release time of the dynamic processor, and the sound image position, width, and depth parameters of the spatial processor. All parameters are optimized based on the feature coordination index of the audio track to ensure good sound quality performance of the processed audio.

[0047] S102. Analyze the frequency distribution law of each audio input signal based on the time-frequency feature map, extract the fundamental frequency and overtone structure features of each audio input signal, and establish a timbre feature fingerprint library according to the fundamental frequency and overtone structure features.

[0048] Among them, the frequency distribution law represents the distribution characteristics of the spectral energy of the audio signal over different frequency components; the fundamental frequency refers to the lowest fundamental frequency component in the audio signal; the harmonic structure feature refers to the frequency and amplitude relationship of each harmonic component in the signal; the timbre feature fingerprint library is a database that stores the feature templates of different sound sources.

[0049] This step is executed after obtaining the time-frequency feature map. Specifically, the system first performs peak detection on the time-frequency feature map to locate the main frequency components. Then, methods such as autocorrelation analysis are used to determine the fundamental frequency of the signal, and each harmonic component is identified based on the integer multiple relationship of the fundamental frequency. The system analyzes the amplitude ratio relationship and time-varying characteristics of each harmonic, extracts the parameters that can characterize the timbre features, and organizes these parameters into a feature vector and stores it in the fingerprint library.

[0050] In some embodiments, feature extraction and fingerprint library construction can be achieved in the following ways: Optionally, the cepstrum analysis method is used to extract the fundamental frequency, the harmonic-noise decomposition is used to extract the harmonic structure, and the timbre feature fingerprint is constructed by combining the energy envelope feature; Optionally, the deep learning method is used to directly learn the timbre feature representation from the time-frequency map, and the convolutional neural network and other models are used to automatically extract features and establish the fingerprint library. It can be understood that other feature extraction methods can also be used to construct the timbre feature fingerprint library, which is not limited here.

[0051] S103. Obtain a preset standard audio signal, extract a multi-dimensional audio feature vector from the preset standard audio signal, and match the multi-dimensional audio feature vector with the timbre feature fingerprint library to obtain an audio feature parameter group.

[0052] Among them, the preset standard audio signal refers to a noise-free and high-quality audio signal recorded in an ideal environment; the multi-dimensional audio feature vector includes parameters such as spectral envelope, energy distribution, transient features, and spatial information; the timbre feature fingerprint library is a data set containing different types of audio feature templates; the audio feature parameter group refers to a complete parameter set for subsequent audio processing.

[0053] This step is executed after establishing the timbre feature fingerprint library. Specifically, the system first selects a standard audio signal with a style similar to the current input signal from the preset standard audio database as a reference sample. Then, time-frequency analysis is performed on the standard audio signal to extract multi-dimensional feature parameters including spectral centroid, spectral flux, harmonic ratio, timbre brightness, etc., to form a feature vector. The system performs a similarity match between the feature vector and the templates in the established timbre feature fingerprint library, selects the most matching feature template, and combines the features of the current signal to generate a feature parameter group containing parameters such as timbre, dynamics, and space.

[0054] In some embodiments, the feature extraction and matching process can be implemented in various ways: Optionally, first perform multi-scale decomposition on the standard audio signal, extract the energy distribution features of different frequency bands respectively, then calculate the statistical features of the signal such as mean, variance, etc., and finally extract the dynamic features of the signal such as envelope features, and combine these features to form a feature vector for matching; Optionally, adopt a machine learning method, automatically learn the representation of audio features by training a deep neural network, the input of the network is the time-frequency map of the audio signal, the output is the corresponding feature vector, and then perform feature matching through similarity calculation. It can be understood that other feature extraction and matching methods can also be adopted to obtain the audio feature parameter group, which is not limited here.

[0055] Taking a concert as an example, the system obtains a preset standard audio signal during the rehearsal stage before the official performance. This signal is collected in a professional recording studio environment (background noise is lower than -80dBFS), and includes the lead singer's vocals (-15dBFS, 100Hz - 10kHz), instrument accompaniment (-20dBFS, 20Hz - 20kHz), and ambient reverberation (-40dBFS, RT60 is 1.2 seconds). The system extracts feature vectors from this standard audio: First, perform spectral analysis through a 1024-point FFT to obtain the energy distribution curves of 31 frequency bands; calculate the spectral centroid (3.2kHz) and spectral skewness (-2.5dB / octave); extract envelope features including attack time (5ms), release time (150ms), and dynamic range (90dB); analyze spatial features to obtain the stereo width (0.8) and depth (0.7) parameters. The system matches these feature vectors with a timbre feature fingerprint library, which contains more than 1000 preset audio feature templates. Find the closest feature template by calculating the Euclidean distance, and extract the corresponding audio feature parameter group: the settings of a 31-band parametric equalizer (center frequency, Q value, and gain), dynamic processor parameters (compression ratio 2:1, threshold -18dB, attack time 5ms, release time 150ms), and spatial processing parameters (stereo correlation 0.85, stereo position ±30°). These parameters extracted from the standard audio will be used as the reference benchmark for subsequent real-time audio processing. The system realizes precise audio optimization processing by comparing the feature vectors of the real-time audio with these standard parameters.

[0056] S104. Construct an audio feature coordination index system according to the audio feature parameter group.

[0057] Among them, the audio feature coordination index system represents a multi-dimensional evaluation system for evaluating the overall sound quality of an audio signal; the overall sound coordination degree refers to the balance index of the energy distribution of each frequency band of the audio signal; the clarity of voice parts represents the distinguishable degree between different voice parts; the spatial stereo balance degree refers to the rationality index of stereo image positioning in a stereo signal.

[0058] This step is executed after obtaining the audio feature parameter group. Specifically, based on the spectrum, dynamics, and spatial parameters in the audio feature parameter group, the system constructs an evaluation index system with multiple dimensions. By analyzing parameters such as the frequency band energy distribution, voice part loudness relationship, and channel signal characteristics, the system establishes a quantitative standard for evaluating the overall quality of the audio signal. The system combines these parameters according to preset weights to form a comprehensive index system that can be used to guide subsequent signal processing.

[0059] In some embodiments, the construction of the index system can be achieved in various ways: Optionally, first establish an evaluation model based on the spectrum energy distribution, calculate the energy ratio of adjacent frequency bands, then analyze the loudness level relationship of the voice part signals, and finally construct a spatial evaluation index by combining the phase and amplitude characteristics of the channel signals; Optionally, adopt a data-driven method, and establish an evaluation system based on a statistical model by analyzing the characteristic distribution laws of a large number of high-quality audio samples, including an energy distribution model, a voice part relationship model, and a spatial characteristic model. It can be understood that other methods can also be used to construct the audio feature coordination index system, which is not limited here.

[0060] This step specifically includes: Calculate the energy values of different frequency bands in the audio feature parameter group, obtain the energy ratio between adjacent frequency bands, and calculate the overall sound coordination degree index according to the energy ratio.

[0061] In this step, the energy value of the frequency band refers to the root mean square energy value of 31 frequency bands; the energy ratio represents the energy comparison relationship between adjacent frequency bands; the overall sound coordination degree index represents the balance degree of the spectrum energy distribution, with a value range of 0 - 1. The system executes the spectrum energy analysis process: First, perform a 31-segment frequency analysis on the audio signal, with the frequency band division using a 1 / 3 octave interval, covering the range of 20 Hz - 20 kHz. Calculate the energy value of each frequency band through a 2048-point FFT, and use the exponential averaging method (average time constant 125 ms) to obtain a stable energy reading. For the energy Ei of the i-th frequency band, calculate its energy ratio with the adjacent frequency bands: the upper ratio Ri_up = Ei / Ei+1, and the lower ratio Ri_down = Ei / Ei-1. Calculate the frequency band energy gradient based on these ratios: Gi = (Ri_up + Ri_down) / 2. When Gi falls within the range of 0.8 to 1.25, this frequency band is considered to have coordinated energy with the adjacent frequency bands. The system counts the coordination status of all frequency bands, calculates the proportion of the number of coordinated frequency bands to the total number of frequency bands, and obtains the overall coordination degree index. At the same time, the system performs variance analysis on the energy ratio sequence, and a variance value less than 0.5 indicates a stable spectrum energy distribution. The final overall coordination degree index is calculated by weighted average: coordination degree = 0.6 × proportion of coordinated frequency bands + 0.4 × (1 - variance value).

[0062] Extract the vocal loudness value and frequency value in the audio feature parameter group, and calculate the vocal layer clarity index corresponding to the overall sound coordination index.

[0063] In this step, the vocal loudness value refers to the loudness level of each vocal part; the frequency value refers to the frequency center and bandwidth of the vocal part; the vocal layer clarity index represents the distinguishable degree of different vocal parts in terms of loudness and frequency. The system performs a vocal analysis process: uses a loudness calculation model (ISO532B) to calculate the loudness level of each vocal part, and at the same time extracts the frequency characteristics of the vocal part. For N vocal parts, construct an N×N vocal part discrimination matrix D. The matrix element Dij represents the discrimination degree between the i-th vocal part and the j-th vocal part, and the calculation formula is: Dij = α×|Li - Lj| / Lmax + β×|Fi - Fj| / Fmax, where Li and Lj are loudness values, Fi and Fj are frequency center values, and α = 0.6 and β = 0.4 are weight coefficients. The diagonal elements of the vocal part discrimination matrix are set to 1, indicating complete self-discrimination. The system calculates the average value of the non-diagonal elements of the matrix to obtain the overall vocal part discrimination degree. When the loudness difference between adjacent vocal parts is greater than 6 dB and the frequency interval is greater than the critical bandwidth, the vocal parts are considered to have a good sense of layering. The system obtains the vocal layer clarity index through the weighted combination of the vocal part discrimination degree and the overall coordination degree: clarity = 0.7×vocal part discrimination degree + 0.3×overall coordination degree.

[0064] According to the left and right channel signal values in the audio feature parameter group, calculate the phase difference and amplitude difference of the channel signals, and determine the sound image position distribution in combination with the vocal layer clarity index to obtain the spatial sound image balance index.

[0065] In this step, the phase difference refers to the time delay between the left and right channel signals; the amplitude difference refers to the energy difference between channels; the sound image position distribution represents the spatial localization of the sound source in the stereo field; the spatial sound image equalization index reflects the uniformity of the sound image distribution. The system performs a spatial sound image analysis process: calculating the phase difference between the left and right channels through short-time cross-correlation analysis, with a time window length of 20 ms and a step of 10 ms. For the frequency component f, the phase difference φ(f) is converted into the sound image azimuth angle θ(f)=arcsin(φ(f)×f / c), where c is the speed of sound. The sound image intensity is obtained by calculating the energy ratio of the left and right channels: I(f)=10×log(EL(f) / ER(f)). The system divides the frequency range of 20 Hz - 20 kHz into 24 critical bands in the frequency domain, and calculates the sound image position vector for each band: P(k)=[θ(k), I(k)]. The sound image distribution map is obtained by statistically accumulating the energy at different azimuth angles, ranging from -30° to +30°. The system calculates the standard deviation σ and skewness S of the sound image distribution to evaluate the uniformity of the spatial distribution. Combining with the voice part level clarity index C, the final spatial sound image equalization is calculated: E=w1×(1-σ / 30)+w2×(1-|S|)+w3×C, where the weight coefficients are w1 = 0.4, w2 = 0.3, and w3 = 0.3. When E is greater than 0.8, it indicates that the sound image distribution has good spatial equalization.

[0066] S105. Calculate the audio feature coordination index of the multi-channel audio input signals in real time based on the audio feature coordination index system, compare the audio feature coordination index with the coordination index of the preset standard audio signal, and obtain the signal adjustment value.

[0067] Among them, real-time calculation means continuous index calculation and update during the continuous input of the audio signal; the audio feature coordination index refers to the quantitative evaluation value calculated based on the index system, including the overall sound coordination index, the voice part level clarity index, and the spatial sound image equalization index; the coordination index of the preset standard audio signal represents the evaluation value of the high-quality reference audio under the same index system; the signal adjustment value refers to the quantized value of the parameter to be adjusted obtained by index comparison, including the equalization adjustment value, the dynamic adjustment value, and the spatial adjustment value.

[0068] This step is continuously executed after establishing the audio feature coordination index system. Specifically, the system first segments the input multi-channel audio signals according to a preset time window, and calculates the index values of three dimensions: the overall sound coordination degree, the clarity of voice parts, and the spatial sound image balance degree for each segment of the signal. During the calculation process, the system analyzes the frequency band energy distribution characteristics to calculate the coordination index, calculates the clarity index through the voice part loudness and frequency characteristics, and calculates the balance index based on the phase and amplitude relationship of the channel signals. Then the system extracts the index values of the corresponding time period from the preset standard audio signal, makes multi-dimensional comparisons between the calculated index values and the standard index values, calculates the difference values of each dimension through the set thresholds and weights, and finally determines the specific value of the signal adjustment according to the matching result of the difference value and the timbre characteristics.

[0069] In some embodiments, the real-time index calculation and comparison process can be implemented in various ways: Optionally, first perform adaptive segmentation on the input signal, dynamically adjust the analysis window length according to the change characteristics of the signal, then use the weighted average method to calculate the index values of each dimension, then use the comparison algorithm based on fuzzy logic for index comparison, and finally predict the optimal adjustment value through a neural network model; Optionally, adopt a parallel computing architecture, process the signals of multiple time windows simultaneously, extract features and calculate the index values for each window using the frequency domain decomposition method, then perform index comparison through a recursive optimization algorithm, and finally dynamically optimize the calculation method of the adjustment value based on the historical adjustment effect. It can be understood that other methods can also be used to implement the real-time index calculation and comparison process, which is not limited here.

[0070] This step specifically includes: According to the overall sound coordination degree calculation rule, the voice part clarity calculation rule, and the spatial sound image balance degree calculation rule in the audio feature coordination index system, segment the multi-channel audio input signals at time intervals, and calculate the coordination index of each segment of the signal.

[0071] Among them, the coordination index system includes calculation rules in three dimensions; the time interval is segmented with an analysis window of 20 ms and a step interval of 10 ms; the coordination index of each segment of the signal includes three values: overall coordination degree, level clarity, and spatial balance degree. In this step, the system performs a segmented coordination analysis process. First, the input multi-channel audio signals are segmented, and each analysis window contains 1024 sampling points (sampling rate 48 kHz). For each time window, the system calculates the indexes in three dimensions: (1) Overall coordination degree calculation: The frequency band is divided by a 31-band parametric equalizer, the energy ratio of adjacent frequency bands (Ei / Ei+1) is calculated, the proportion of frequency bands where the energy ratio falls within the range of 0.8 - 1.25 is statistically calculated, and the variance of the spectral energy distribution is calculated to obtain the overall coordination degree Ct; (2) Level clarity calculation: The loudness level Li and the center frequency Fi of each voice part are extracted, an N×N voice part discrimination matrix is constructed, and the matrix element Dij = 0.6×|Li - Lj| / Lmax + 0.4×|Fi - Fj| / Fmax. The average value of the non-diagonal elements is calculated to obtain the level clarity Lt; (3) Spatial balance degree calculation: The phase difference φ(f) and the energy ratio I(f) of the left and right channels are analyzed. The sound image position vector P(k) = [θ(k), I(k)] is calculated in 24 critical frequency bands. The spatial balance degree St = 0.4×(1 - σ / 30) + 0.3×(1 - |S|) + 0.3×Lt is calculated through the standard deviation σ and skewness S of the sound image distribution. The system combines these three indexes into a coordination index vector [Ct, Lt, St] as the feature representation of this time window.

[0072] Extract the coordination indexes of the corresponding time period from the preset standard audio signal, perform a numerical comparison with the coordination indexes of the multi-channel audio input signal, and obtain the numerical comparison result.

[0073] In this step, among them, the preset standard audio signal refers to the ideal audio recorded in a noise-free environment; the corresponding time period refers to the time segment with the same music content; the numerical comparison includes the difference calculation of each dimension index. The system performs an index comparison process. For each 20-ms analysis window, the system locates the corresponding time position in the standard audio and extracts the coordination index vector [Cs, Ls, Ss] of the standard audio. The system calculates the differences between the input signal and the standard signal in three dimensions: the overall coordination degree difference ΔC = |Ct - Cs|, the level clarity difference ΔL = |Lt - Ls|, and the spatial balance degree difference ΔS = |St - Ss|. For any time window i, the system calculates the weighted difference value Di = 0.4×ΔC + 0.35×ΔL + 0.25×ΔS. When Di is greater than the threshold 0.2, this time window is marked as needing adjustment. The system also records the specific difference directions and values of each dimension to provide a basis for subsequent parameter adjustment.

[0074] Determine the signal adjustment value according to the numerical comparison result and the matching features of the timbre feature fingerprint library.

[0075] In this step, the numerical comparison result includes the difference values and directions of each dimension; the timbre feature fingerprint library contains standardized adjustment parameter templates; the signal adjustment value refers to the specific parameter settings of each processing module. The system executes the parameter adjustment calculation process. For each time window that needs to be adjusted, the system first searches for the most matching feature template in the timbre feature fingerprint library. The matching process uses the Euclidean distance metric, and the feature template with the smallest distance is selected as the reference. The system calculates the specific adjustment parameters according to the difference analysis result: (1) Frequency response adjustment: When ΔC shows spectral imbalance, calculate the gain adjustment value gi = -10×log(Ei / Es,i) of the 31-band equalizer according to the energy ratio, where Ei and Es,i are the band energies of the current signal and the standard signal respectively; (2) Dynamic range adjustment: When ΔL shows insufficient layering, calculate the compressor parameters: compression ratio r = 1 + 2×ΔL, threshold th = -18 - 12×ΔL dB; (3) Spatial processing adjustment: When ΔS shows uneven spatial distribution, calculate the pan adjustment parameters: azimuth offset θadj = -30×ΔS×sign(S) degrees, width adjustment wadj = 0.5×ΔS. The system combines these adjustment parameters into a complete processing instruction and applies it to the audio signal processing.

[0076] S106. Calculate the target audio processing parameter combination based on the signal adjustment value and in combination with the timbre feature fingerprint library.

[0077] Among them, the signal adjustment value represents the specific numerical parameters for adjusting the audio signal; the matching features of the timbre feature fingerprint library refer to the parameter information contained in the feature template that is closest to the current audio signal; the target audio processing parameter combination represents the set of parameters for optimizing the audio signal, including equalization parameters, dynamic parameters, and spatial parameters; the equalization parameters are used to adjust the frequency response characteristics; the dynamic parameters are used to control the dynamic range of the signal; the spatial parameters are used to optimize the stereo pan effect.

[0078] This step is executed after obtaining the signal adjustment value. Specifically, the system first analyzes the correlation between the signal adjustment value and the matching features in the timbre feature fingerprint library and establishes a parameter mapping model. Then, according to the adjustment requirements of different frequency bands and in combination with the time-varying characteristics of the timbre features, calculate the equalization parameters of each frequency band, including the center frequency, gain, and Q value. At the same time, the system calculates the dynamic processing parameters such as the compression ratio and threshold based on the adjustment requirements of the dynamic range and in combination with the transient characteristics of the signal. In addition, the system also needs to calculate the stereo processing parameters, including the pan position, width, and depth parameters, according to the adjustment requirements of the spatial pan and in combination with the correlation of the channel signals.

[0079] In some embodiments, the calculation process of parameter combinations can be achieved in various ways: Optionally, first construct a parameter optimization model based on an expert system, establish a rule base according to historical processing experience, then determine the initial parameter values through rule inference, then use a genetic algorithm to optimize the parameter combination, and finally ensure the stability of the parameters through cross-validation; Optionally, adopt a deep reinforcement learning method, model the parameter adjustment process as a Markov decision process, continuously optimize the parameter selection strategy by interacting with the environment, and at the same time combine a multi-objective optimization algorithm to balance different processing objectives, and finally obtain the optimal parameter combination. It can be understood that other methods can also be used to implement the calculation of the target audio processing parameter combination, which is not limited here.

[0080] The following further describes the method provided in this embodiment in a more specific process. Please refer to Figure 2 , which is another schematic flowchart of the digital audio signal processing method of the mixing console in the embodiment of the present application.

[0081] S201. Based on the signal adjustment value, calculate the target audio processing parameter combination in combination with the timbre feature fingerprint library. The target audio processing parameter combination includes equalization parameters, dynamic parameters, and spatial parameters.

[0082] Among them, the signal adjustment value represents the parameter value to be adjusted obtained after feature comparison; the timbre feature fingerprint library stores pre-established timbre feature templates; the equalization parameters include the center frequency, Q value, and gain value of each frequency band; the dynamic parameters include compression ratio, threshold, attack time, and release time; the spatial parameters include pan position, width, and depth parameters.

[0083] The system first extracts the feature template that best matches the current signal from the timbre feature fingerprint library. This template contains preset timbre types, spectral envelopes, and dynamic features. Based on the extracted feature template, the system calculates the target frequency response curve of each frequency band, specifically by performing 1 / 3 octave analysis on the spectrum and calculating the target gain values of 30 frequency bands in the range of 31.5 Hz - 16 kHz. For the dynamic parameters, the system determines the compression threshold and compression ratio by calculating the peak level and root mean square level of the signal, and sets the compression curve in the range of -20 dB to 0 dB. For the spatial parameters, the system analyzes the phase difference and level difference between the left and right channels and calculates the pan position parameters in the range of ±60°. The system combines these parameters according to the preset priority to generate a target processing parameter combination containing 93 specific parameters.

[0084] S202. Analyze the timbre feature distribution in the time-frequency feature map, and identify the signal features of different musical instruments and vocals according to the timbre feature distribution.

[0085] Among them, the time-frequency feature map represents the spectral features of the audio signal changing with time; the timbre feature distribution refers to the energy distribution pattern of different sound sources on the time-frequency plane; the signal features include the fundamental frequency, harmonic structure, and transient features.

[0086] The system performs adaptive segmented analysis on the time-frequency feature map. Each segment uses a Hamming window of 32 ms with a step size of 16 ms. By calculating the spectral centroid, spectral flux, and sub-band energy ratio, the timbre feature parameters of each time window are extracted. The system uses Mel Frequency Cepstral Coefficients (MFCC) to extract 20 timbre feature coefficients and combines the pitch detection algorithm to identify the pitch of each voice part. By analyzing the harmonic structure, the system identifies the characteristic frequency components of different musical instruments. For example, the piano has characteristic harmonics of 2096 Hz and 3144 Hz at the fundamental frequency of 262 Hz (C4). For human voice signals, the system identifies vowel features through formant tracking and combines zero-crossing rate analysis to identify consonant segments. The system establishes a multi-dimensional discriminant model based on these features to achieve feature classification of different sound sources.

[0087] S203. Separate the multi-channel audio input signals into independent audio tracks according to the signal features.

[0088] Among them, the independent audio track represents the single sound source signal after separation; signal separation refers to the process of extracting different sound source components from the mixed signal.

[0089] The system adopts a deep learning-based sound source separation algorithm and uses a convolutional neural network with a U-Net structure to separate the signal. The input layer receives the time-frequency feature map and extracts hierarchical features through a 5-layer encoder-decoder structure. Each encoder layer contains two 3×3 convolutional layers and a max-pooling layer, and the decoder uses transposed convolution for upsampling. The skip-connection structure preserves the detailed features. The system applies Wiener filtering to the spectrogram output by the network to reconstruct the time-domain signal. The processing uses a 2048-point FFT with 50% overlap, and the output sampling rate is 48 kHz for the independent audio track. The system performs phase correction on the separated signal to eliminate the phase distortion introduced during the separation process and maintain the time alignment relationship of the original signal. The finally output independent audio track maintains the dynamic range and frequency characteristics of the original signal.

[0090] S204. Extract the fundamental frequency and harmonic structure features of the independent audio track and perform feature matching with the timbre feature fingerprint library.

[0091] Among them, the fundamental frequency represents the basic frequency component of the pitch; the harmonic structure features include the frequency and energy distribution of each harmonic; feature matching refers to the process of calculating the similarity between the extracted features and the preset template.

[0092] The system performs autocorrelation analysis on each independent audio track, using a Hamming window with a frame length of 1024 points, and calculates the fundamental frequency through the autocorrelation function. For vocal signals, the system searches for the fundamental frequency in the range of 70 - 400 Hz; for instrument signals, the system expands the search range to 20 - 2000 Hz. The system extracts the first 16 harmonic components through the harmonic-noise decomposition algorithm, calculates the energy ratio of each harmonic relative to the fundamental frequency to form a harmonic feature vector. At the same time, the system calculates the frequency ratio of adjacent harmonics to identify the change characteristics of the harmonic structure. The system calculates the cosine similarity between the extracted feature vector and the templates in the timbre feature fingerprint library, and selects the feature template with a similarity exceeding 0.85 as the matching result. When a significant change in the harmonic structure is detected (the change in the energy ratio of adjacent harmonics exceeds 6 dB), the system updates the feature matching result.

[0093] S205. Calculate the audio feature coordination index of the independent audio track to obtain the target parameter combination of each track.

[0094] Among them, the audio feature coordination index includes spectral balance, dynamic consistency, and spatial correlation; the target parameter combination includes equalization, compression, and spatial processing parameters.

[0095] The system calculates the coordination index of each independent audio track respectively. In terms of spectral balance, the system divides the frequency range into five sub-bands: low frequency (20 - 200 Hz), mid-low frequency (200 - 800 Hz), mid-frequency (800 - 3 kHz), mid-high frequency (3 - 8 kHz), and high frequency (8 - 20 kHz), and calculates the energy ratio of adjacent sub-bands and the change trend of the spectral centroid. The dynamic consistency is quantified by calculating the change rate of the peak factor (PeakFactor) and the root mean square level (RMS) of the signal. The system sets the target range of the peak factor to 8 - 12 dB. The spatial correlation is evaluated by calculating the cross-correlation function and the phase difference spectrum of the stereo signal to determine the optimal sound image position within the range of ±45°. Based on these indexes, the system generates a target parameter combination including 31-band equalization parameters, 4 groups of dynamic processing parameters, and 3 spatial processing parameters.

[0096] S206. Extract the frequency change characteristics and energy change characteristics in the time-frequency feature map, and analyze the change rules of the characteristics.

[0097] Among them, the frequency change characteristics represent the change of the signal frequency components over time; the energy change characteristics refer to the time-varying characteristics of the signal energy distribution; the change rules refer to the periodic or trend performance of these characteristics.

[0098] The system conducts continuous analysis on the time-frequency feature map, using a 1024-point FFT and a 75% frame overlap rate, with a time resolution of 5.33 ms. By calculating the spectral centroid and spectral skewness of each frame of the signal, the system obtains the frequency change curve. The system uses the time series of the sub-band energy ratio to describe the energy change characteristics, and establishes an energy envelope model by calculating the mean and standard deviation of 20 adjacent frames. To extract the long-term change pattern, the system performs wavelet decomposition on the feature sequence and uses a 5-level Daubechies wavelet for multi-resolution analysis. The system fits the feature change trend through an autoregressive model and calculates the periodicity of the frequency change (such as the rhythm feature) and the envelope feature of the energy change (such as the intensity change). For mutation features, the system detects and marks them by setting a relative threshold of 0.5.

[0099] S207. Predict the fundamental frequency change trend and the harmonic structure change trend in the next time period based on the change pattern and the historical data of the audio feature parameter group.

[0100] Among them, the change pattern refers to the frequency and energy change characteristics extracted from the time-frequency feature map; the historical data of the audio feature parameter group refers to the feature parameter records within the past 100 ms; the fundamental frequency change trend refers to the change direction and amplitude of the fundamental frequency within the next 50 ms; the harmonic structure change trend includes the expected changes in the harmonic frequency ratio and the energy ratio.

[0101] The system uses a long short-term memory network (LSTM) to perform predictive analysis on the feature sequence. The input layer receives the feature sequence of 20 time frames, including the fundamental frequency trajectory and the energy ratios of 16 harmonic components. The LSTM network contains two hidden layers, with 128 neurons in each layer, and uses the tanh activation function. The system normalizes the input features, standardizes the fundamental frequency to the range of 0 - 1, and compresses the harmonic energy ratio to the range of -60 dB to 0 dB through logarithmic transformation. The network training uses the root mean square error as the loss function and uses the Adam optimizer to update the parameters. The system obtains the feature changes of the next 10 time frames through recursive prediction and applies the Kalman filter to smooth the prediction results to obtain a continuous change trend curve.

[0102] S208. Adjust the numerical values of each index in the audio feature coordination index system based on the fundamental frequency change trend and the harmonic structure change trend.

[0103] Among them, the index numerical values include the quantization values of the overall sound coordination degree, the clarity of the voice part levels, and the spatial sound image balance degree; the index adjustment refers to the process of correcting the index calculation parameters according to the predicted trend.

[0104] The system establishes an index adjustment model based on a deep neural network. This model adopts a three-layer fully connected structure, and the input layer receives the predicted trend feature vector. The first hidden layer contains 256 neurons and uses the ReLU activation function to process the fundamental frequency change features; the second hidden layer contains 128 neurons and processes the harmonic structure change features. The system divides the overall coordination index into sub-indices of 5 frequency bands and adjusts the weight coefficients of each sub-index according to the predicted frequency changes. The clarity index of the voice part adapts to signal changes by dynamically adjusting the detection threshold (range from -40 dB to -20 dB). The spatial sound image balance index dynamically adjusts the phase correlation determination threshold (range from 0.6 to 0.9) according to the predicted phase relationship changes. The system applies exponential smoothing to the adjusted index values to ensure the continuity of index changes.

[0105] S209. Substitute the adjusted numerical values of each index into the calculation process of the target audio processing parameter combination to obtain a pre-adjusted parameter group including equalization parameters, dynamic parameters, and spatial parameters.

[0106] Among them, the calculation process of the target audio processing parameter combination refers to the mapping method that converts the index values into specific processing parameters; the pre-adjusted parameter group includes equalization parameters of 31 frequency bands, 4 groups of dynamic processing parameters, and 3 spatial processing parameters.

[0107] The system uses the gradient descent algorithm to optimize the parameter mapping process, and the objective function combines evaluation indices in three dimensions: equalization effect, dynamic control, and spatial positioning. The calculation of the equalization parameters uses 31 1 / 3 octave filters with center frequencies ranging from 20 Hz to 20 kHz and a fixed Q value of 4.32. The system calculates the gain value of each filter according to the overall coordination index, and the range is limited to ±12 dB. The dynamic parameters include compression ratio (1:1 to 10:1), threshold (-40 dB to 0 dB), attack time (0.1 ms to 50 ms), and release time (10 ms to 1000 ms). The spatial parameters are calculated by parameters in three dimensions: sound image position (-60° to +60°), width (0 to 1), and depth (0 to 1). The system performs clipping and smoothing processing on the calculated parameters to ensure the smoothness of parameter changes.

[0108] Taking the processing of a human voice signal as an example, the system executes the parameter calculation process. In the equalization parameter calculation step, the system detects that the coordination index of the human voice signal in the 200 - 800 Hz frequency band is 0.75 (the standard value is 0.85), and the clarity index in the 2 kHz - 4 kHz frequency band is 0.65 (the standard value is 0.80). Based on these index values, the system sets the center frequency to 400 Hz and the Q value to 4.32 in the 200 - 800 Hz frequency band, and calculates the +3 dB gain value through the formula: gain = 10×log(0.85 / 0.75); in the 2k - 4 kHz frequency band, it sets the center frequency to 3 kHz and the Q value to 4.32, and calculates the +4.5 dB gain value through the formula: gain = 10×log(0.80 / 0.65). In the dynamic parameter calculation step, the system detects that the dynamic range index of the signal is 0.60 (the standard value is 0.75), and the peak factor is 15 dB (the target value is 12 dB). Based on this, the compression ratio is calculated to be 2.5:1 (calculation formula: 1+(15 - 12) / 2), the compression threshold is set to -18 dB (set based on the peak level - 3 dB), the attack time is set to 20 ms, and the release time is set to 150 ms. In the spatial parameter calculation step, the system detects that the stereo correlation index is 0.85 (the standard value is 0.90), and the sound image localization index is 0.70 (the standard value is 0.85). Based on this, the sound image position adjustment value is calculated to be ±15° (calculated based on the correlation difference), the sound image width is 0.8 (calculated through the formula: width = 0.85 / 0.90), and the sound image depth is 0.7 (calculated through the formula: depth = 0.70 / 0.85). The system combines these calculated parameters into a pre-adjustment parameter group for subsequent real-time audio processing. The parameter calculation process uses recursive averaging, that is, each newly calculated parameter value is mixed with the previous parameter value in a ratio of 0.7:0.3 to ensure the smoothness of parameter changes. Rate limits are set for all parameters. The maximum change rate of the equalization gain is 6 dB / s, the change of dynamic parameters uses logarithmic gradual change, and the change of spatial parameters uses linear interpolation.

[0109] S210. Obtain an environmental noise signal, convert the environmental noise signal into a noise time-frequency feature map, and extract the frequency distribution feature and energy distribution feature in the noise time-frequency feature map.

[0110] Among them, the environmental noise signal represents the background noise collected by a microphone array; the noise time-frequency feature map refers to the time-frequency analysis result of the noise signal; the frequency distribution feature includes the energy distribution curve of the noise in the frequency domain; the energy distribution feature represents the change feature of the noise intensity over time.

[0111] The system collects ambient noise through an 8-channel circular microphone array, with the sampling rate set at 48 kHz and the quantization precision at 24 bits. The collected noise signal is subjected to short-time Fourier transform (STFT), and a time-frequency feature map is generated using a 2048-point FFT and a 50% frame overlap rate. The system divides the frequency bands of the time-frequency feature map, equally divides the range of 20 Hz - 20 kHz into 24 sub-bands, and calculates the energy spectral density of each sub-band. By calculating the energy difference between adjacent time frames, the system obtains the time-varying characteristics of the noise and establishes an energy envelope model. For the frequency distribution characteristics, the system calculates the mean and variance of each sub-band to form a noise spectrum feature vector. The system quantifies the frequency concentration and spectral shape characteristics of the noise by calculating the spectral centroid and spectral flatness. At the same time, the system calculates the temporal correlation of the energy of each sub-band to identify the periodic components of the noise.

[0112] S211. Compare the frequency distribution characteristics and energy distribution characteristics of the noise time-frequency feature map with those of the audio input signal time-frequency feature map to determine the noise frequency band and noise energy.

[0113] Among them, the noise frequency band refers to the frequency range where the noise energy is concentrated; the noise energy represents the intensity level of the noise in each frequency band; the comparison process refers to identifying the noise components through feature similarity analysis.

[0114] The system uses a two-dimensional cross-correlation analysis method to compare the time-frequency feature maps. First, the two time-frequency feature maps are resampled at the same time-frequency resolution to ensure that the time-frequency points correspond one by one. The system calculates the correlation coefficient of each time-frequency point and constructs a correlation matrix. By setting a correlation threshold of 0.7, the system identifies the time-frequency regions with high correlation, and these regions are marked as noise-dominated regions. For each frequency band, the system calculates the signal-to-noise ratio (SNR), and marks the frequency bands with an SNR lower than 6 dB as noise frequency bands. The system estimates the noise energy level of each frequency band according to the least mean square error criterion and establishes a noise power spectrum model. At the same time, the system distinguishes between steady-state noise and impulsive noise components by analyzing the fluctuation characteristics of the energy envelope.

[0115] S212. Construct a noise suppression index based on the noise frequency band and noise energy, and add the noise suppression index to the audio feature coordination index system.

[0116] Among them, the noise suppression index includes a spectral masking index and an energy suppression index; the spectral masking index represents the degree of spectral interference of the noise on the effective signal; the energy suppression index represents the noise energy level that needs to be reduced.

[0117] The system constructs a noise suppression index based on the principle of auditory masking. First, the masking threshold curves of 24 critical bands are established, and the critical band widths are divided according to the Bark scale. The system calculates the signal-to-noise ratio and audibility of each band to generate a spectral masking index. When the signal energy exceeds the masking threshold by 12 dB, the band is marked as a signal band that needs to be protected. The system designs a suppression curve according to the noise energy level, with an attenuation slope of -12 dB / octave for steady-state noise and -18 dB / octave for impulsive noise. The system adds the constructed noise suppression index to the existing index system and combines it with other indexes through a weight coefficient of 0.3. The new index participates in the subsequent parameter optimization process to guide the system to perform noise suppression processing.

[0118] Taking a voice recording in a recording studio environment as an example, the system executes the construction process of the noise suppression index. First, the system detects that the energy level of the air conditioner noise in the 100 - 300 Hz band is -45 dBFS, the energy level of the environmental resonance at 500 Hz is -50 dBFS, and the energy level of the high-frequency equipment noise in the band above 8 kHz is -60 dBFS. Based on these noise characteristics, the system constructs a masking threshold curve on 24 critical bands. In the 100 - 300 Hz band, since the signal energy is -35 dBFS and the signal-to-noise ratio is 10 dB, the system sets the spectral masking index to 0.8; in the 500 Hz band, the signal energy is -30 dBFS and the signal-to-noise ratio is 20 dB, and the spectral masking index is set to 0.9; in the band above 8 kHz, the signal energy is -40 dBFS and the signal-to-noise ratio is 20 dB, and the spectral masking index is set to 0.85. For the energy suppression index, the system sets an attenuation curve of -12 dB / octave in the low-frequency band, -9 dB / octave in the middle-frequency band, and -15 dB / octave in the high-frequency band. These indexes are added to the existing audio feature coordination index system through weighted average (weight coefficient is 0.3), where the weight of the low-frequency band is 0.4, the weight of the middle-frequency band is 0.4, and the weight of the high-frequency band is 0.2. The system sets a dynamic update mechanism for the newly added index, updates the index value every 50 ms, and ensures the continuity of the index change through exponential smoothing (smoothing coefficient is 0.8). The noise suppression index acts together with the original spectral balance index, dynamic consistency index, and spatial correlation index to form a more complete audio quality evaluation system.

[0119] S213. Adjust the equalization parameters and dynamic parameters in the target audio processing parameter combination based on the value of the noise suppression index.

[0120] Among them, the numerical values of the noise suppression index include the spectral masking values and energy suppression values of 24 critical bands; the equalization parameters refer to the center frequencies, bandwidths, and gain values of a 31-band parametric equalizer; the dynamic parameters include the threshold, ratio, attack time, and release time of the compressor; the adjustment process represents the calculation method for modifying the audio processing parameters according to the noise characteristics.

[0121] The specific process of the system performing parameter adjustment is as follows: First, the 31-band parametric equalizer is adjusted. The system extends the noise suppression index of 24 critical bands to 31 frequency bands through cubic spline interpolation. For each frequency band, the system calculates the compensation gain according to the spectral masking value, and the compensation range is limited between -12 dB and +6 dB. When the signal-to-noise ratio of a certain frequency band is lower than 6 dB, the system increases the Q value of that frequency band (adjusted from 4.32 to 8.65) to improve the frequency selectivity. For the dynamic parameters, the system adjusts the compressor parameters according to the energy suppression value. In the frequency bands with high noise energy (signal-to-noise ratio < 0 dB), the compression ratio is increased by 50% from the original value, with a maximum not exceeding 10:1. The compressor threshold is adjusted according to the noise energy level, and the calculation formula is: new threshold = original threshold - (noise level + 6 dB). For impulsive noise, the system shortens the attack time of the compressor to 0.1 ms and extends the release time to 500 ms. For steady-state noise, the attack time is set to 1 ms and the release time is set to 100 ms. The system smooths the adjusted parameters through an adaptive filter to avoid audio distortion caused by parameter mutations. Finally, the system applies the adjusted parameters to the audio processing chain to achieve precise noise suppression.

[0122] Taking a piece of audio recorded in a noisy environment as an example, the system executes the parameter adjustment process. The system detects that there is air conditioner noise in the low frequency band (100 - 300 Hz), with a spectral masking value of 0.8 and an energy suppression value of -12 dB; there is environmental resonance in the middle frequency band (500 Hz), with a spectral masking value of 0.9 and an energy suppression value of -9 dB; there is equipment noise in the high frequency band (above 8 kHz), with a spectral masking value of 0.85 and an energy suppression value of -15 dB. In response to these noise characteristics, the system first adjusts the settings of the 31-band parametric equalizer: in the 100 - 300 Hz band, the Q value is increased from 4.32 to 8.65 to enhance frequency selectivity, and the gain values of 5 equalizers within this band are calculated through cubic spline interpolation, ranging from -8 dB to -4 dB; in the 500 Hz band, a narrowband notch filter is set, with the Q value set to 12.5 and the gain value to -6 dB; in the band above 8 kHz, a shelving filter is adopted, with the slope set to -12 dB / octave. For the dynamic parameters, the system increases the compression ratio from the original 2:1 to 3:1 in the low-frequency noise band, reduces the threshold to -24 dB (original signal level - 6 dB), shortens the attack time to 0.1 ms to quickly respond to noise mutations, and extends the release time to 500 ms to smooth the processing effect; in the middle-frequency resonance band, the compression ratio is set to 4:1, the threshold is set to -20 dB, the attack time is 1 ms, and the release time is 200 ms; in the high-frequency noise band, an expander is used for processing, with a ratio of 1:2, the threshold is set to -45 dB, the attack time is 5 ms, and the release time is 100 ms. The system performs recursive averaging processing on all parameter adjustments, mixes the new parameter values and the original parameter values in a ratio of 0.6:0.4, and limits the parameter change rate: the maximum change rate of the equalizer gain is 8 dB / s, the change of the compression parameter adopts logarithmic gradual change, and the adjustment of the time constant adopts linear interpolation to ensure the smoothness and continuity of the processing process.

[0123] The following describes the mixer digital audio signal processing system in the embodiments of the present invention application from the perspective of hardware processing. Please refer to Figure 3 , which is a schematic structural diagram of an entity device of the mixer digital audio signal processing system in the embodiments of the present application.

[0124] It should be noted that Figure 3 the structure of the mixer digital audio signal processing system shown is only an example, and should not impose any limitations on the functions and usage scopes of the embodiments of the present invention.

[0125] As Figure 3As shown, the digital audio signal processing system of the mixing console includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the Read-Only Memory (ROM) 302 or the program loaded from the storage section 308 into the Random Access Memory (RAM) 303, such as executing the method described in the above embodiments. In the RAM 303, various programs and data required for system operation are also stored. The CPU 301, ROM 302, and RAM 303 are connected to each other via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0126] The following components are connected to the I / O interface 305: an input section 306 including an audio input device, button switches, etc.; an output section 307 including a Liquid Crystal Display (LCD), an audio output device, indicator lights, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read from it can be installed into the storage section 308 as needed.

[0127] Specifically, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication section 309, and / or installed from the removable medium 311. When the computer program is executed by the Central Processing Unit (CPU) 301, various functions defined in the present invention are executed.

[0128] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fibers, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0129] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present invention. Among them, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the above-mentioned module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order from that marked in the accompanying drawings.

[0130] Specifically, the mixer digital audio signal processing system of this embodiment includes a processor and a memory, and a computer program is stored on the memory. When the computer program is executed by the processor, the mixer digital audio signal processing method provided in the above embodiment is implemented.

[0131] On the other hand, the present invention also provides a computer-readable storage medium, which may be included in the mixer digital audio signal processing system described in the above embodiment; or it may exist separately and not be assembled into the mixer digital audio signal processing system. The above storage medium carries one or more computer programs. When the above one or more computer programs are executed by a processor of the mixer digital audio signal processing system, the mixer digital audio signal processing system is enabled to implement the mixer digital audio signal processing method provided in the above embodiment.

[0132] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the various embodiments of the present application.

[0133] As used in the foregoing embodiments, depending on the context, the term "when" can be interpreted to mean "if" or "after" or "in response to determining" or "in response to detecting". Similarly, depending on the context, the phrase "upon determining" or "if (the stated condition or event) is detected" can be interpreted to mean "if determined" or "in response to determining" or "when (the stated condition or event) is detected" or "in response to detecting (the stated condition or event)".

[0134] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the foregoing embodiments can be implemented by a computer program instructing relevant hardware. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the foregoing method embodiments. The foregoing storage medium includes: various media such as ROM or random access memory RAM, magnetic disk, or optical disc that can store program code.

Claims

1. A method for processing digital audio signals of a mixing console, characterized in that, Applied to a digital audio signal processing system of a mixing console, the method includes: Obtain multiple audio input signals, and perform time-frequency domain conversion on the audio input signals to obtain time-frequency feature maps; Based on the time-frequency feature maps, analyze the frequency distribution rules of each audio input signal, extract the fundamental frequency and overtone structure features of each audio input signal, and establish a timbre feature fingerprint library according to the fundamental frequency and the overtone structure features; Obtain a preset standard audio signal, extract a multi-dimensional audio feature vector from the preset standard audio signal, and match the multi-dimensional audio feature vector with the timbre feature fingerprint library to obtain an audio feature parameter group; Construct an audio feature coordination index system according to the audio feature parameter group, and the audio feature coordination index system includes overall sound coordination degree, voice part level clarity, and spatial sound image balance degree; Based on the audio feature coordination index system, calculate the audio feature coordination index of the multiple audio input signals in real time, and compare the audio feature coordination index with the coordination index of the preset standard audio signal to obtain a signal adjustment value; Based on the signal adjustment value, calculate a target audio processing parameter combination in combination with the timbre feature fingerprint library, and the target audio processing parameter combination includes equalization parameters, dynamic parameters, and spatial parameters.

2. The method according to claim 1, characterized in that, The step of constructing an audio feature coordination index system according to the audio feature parameter group specifically includes: Calculate the energy values of different frequency bands in the audio feature parameter group, obtain the energy ratio between adjacent frequency bands, and calculate the overall sound coordination degree index according to the energy ratio; Extract the voice part loudness values and frequency values in the audio feature parameter group, and calculate the voice part level clarity index corresponding to the overall sound coordination degree index; According to the left and right channel signal values in the audio feature parameter group, calculate the phase difference and amplitude difference of the channel signals, and determine the sound image position distribution in combination with the voice part level clarity index to obtain the spatial sound image balance degree index.

3. The method according to claim 1, wherein The step of calculating the audio feature coordination index of the multiple audio input signals in real time based on the audio feature coordination index system and comparing the audio feature coordination index with the coordination index of the preset standard audio signal specifically includes: According to the overall sound coordination degree calculation rule, voice part level clarity calculation rule, and spatial sound image balance degree calculation rule in the audio feature coordination index system, segment the multiple audio input signals at time intervals, and calculate the coordination index of each segment of the signal; Extract the coordination index of the corresponding time period from the preset standard audio signal, and perform a numerical comparison with the coordination index of the multiple audio input signals to obtain a numerical comparison result; Determine the signal adjustment value according to the numerical comparison result and the matching features of the timbre feature fingerprint library.

4. The method according to claim 1, characterized in that, After the step of calculating the target audio processing parameter combination in combination with the timbre feature fingerprint library based on the signal adjustment value, the method further includes: Extract the feature change rule of the preset standard audio signal and establish a feature change record table; Analyze the adjustment direction of the target parameter combination according to the feature change record table, and generate a parameter adjustment step size; Adjust the target audio processing parameter combination according to the parameter adjustment step size.

5. The method according to claim 1, characterized in that, After the step of obtaining a plurality of audio input signals and performing time-frequency domain conversion on the audio input signals to obtain a time-frequency feature map, the method further includes: Analyze the timbre feature distribution in the time-frequency feature map, and identify signal features according to the timbre feature distribution; the signal features include musical instrument sounds or human voices; Separate the plurality of audio input signals into independent audio tracks according to the signal features; Extract the fundamental frequency and overtone structure features of the independent audio tracks, and perform feature matching with the timbre feature fingerprint library; Calculate the audio feature coordination index of the independent audio tracks to obtain the target parameter combination of each audio track.

6. The method according to claim 1, wherein After the step of calculating the target audio processing parameter combination based on the signal adjustment value and combining the matching features of the timbre feature fingerprint library, the method further includes: Extract the frequency change feature and energy change feature in the time-frequency feature map, and determine the change rules of the frequency change feature and the energy change feature; Predict the fundamental frequency change trend and overtone structure change trend in the next time period according to the change rules and the historical data of the audio feature parameter group; Based on the fundamental frequency change trend and overtone structure change trend, adjust the values of each index in the audio feature coordination index system; Substitute the adjusted values of each index into the calculation process of the target audio processing parameter combination to obtain a pre-adjusted parameter group including equalization parameters, dynamic parameters, and spatial parameters.

7. The method according to claim 6, characterized in that, After the step of calculating the target audio processing parameter combination based on the signal adjustment value and combining the timbre feature fingerprint library, the method further includes: Obtain an environmental noise signal, convert the environmental noise signal into a noise time-frequency feature map, and extract the frequency distribution feature and energy distribution feature in the noise time-frequency feature map; Compare the frequency distribution feature and energy distribution feature of the noise time-frequency feature map with the time-frequency feature map of the audio input signal to determine the noise frequency band and noise energy; Construct a noise suppression index according to the noise frequency band and noise energy, and add the noise suppression index to the audio feature coordination index system; Based on the value of the noise suppression index, adjust the equalization parameters and dynamic parameters in the target audio processing parameter combination.

8. A digital audio signal processing system for a mixing console, characterized in that, The mixing console digital audio signal processing system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the mixing console digital audio signal processing system to execute the method according to any one of claims 1-7.

9. A computer-readable storage medium, comprising instructions, characterized in that, When the instruction runs on the mixing console digital audio signal processing system, it enables the mixing console digital audio signal processing system to execute the method according to any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product runs on a mixing console digital audio signal processing system, the mixing console digital audio signal processing system is caused to execute the method according to any one of claims 1 to 7.

Images