Digital audio processing method and system for sound equipment

Abstract

The invention relates to the technical field of digital audio processing, in particular to a digital audio processing method and system for sound equipment. The system comprises an audio signal acquisition module, a signal analysis module, an adaptive adjustment module, a howling suppression module and a sound effect optimization feedback module. The audio signal acquisition module acquires data such as frequency spectrum and amplitude of the sound equipment in real time. And the signal analysis module generates a signal analysis report based on a non-negative matrix factorization algorithm and a generative adversarial network algorithm. And the adaptive adjustment module automatically adjusts audio parameters through a fuzzy logic control algorithm. The howling suppression module suppresses howling noise by using an adaptive feedback cancellation algorithm. And the sound effect optimization feedback module adopts a Bezier curve to enhance the sound effect and realize frequency, phase and space positioning optimization. Through cooperative processing of audio acquisition, accurate analysis, adaptive adjustment, howling suppression and spatial sound effect optimization, the audio quality and stability are improved, and the audio spatial effect and user experience are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of digital audio processing, and in particular to a digital audio processing method and system for audio. Background Art

[0002] In modern audio equipment, the quality of sound directly affects the user's auditory experience. With the rapid development of digital technology, digital audio processing has gradually become an important means to improve the sound quality of audio equipment. Traditional audio systems mainly rely on analog circuits to process audio signals. This method has certain limitations in terms of noise, distortion and frequency response, and it is difficult to meet users' demand for high-fidelity sound quality. In recent years, digital audio processing systems have enhanced audio output effects by applying digital signal processing technologies (such as filtering, equalization, echo cancellation, reverberation, etc.). At present, there are still the following problems: the existing systems mostly rely on fixed algorithms or manual adjustments, and cannot automatically adapt to different environments to achieve the best sound effects; traditional suppression algorithms cannot dynamically eliminate howling noise in real time, which affects the user experience; the existing systems have limited performance in spatial positioning and sound enhancement, and it is difficult to achieve an immersive experience. Summary of the invention

[0003] To solve the above problems, the present invention provides a digital audio processing method and system for audio, which solves the problem of how to design a digital audio processing system in an audio system, which can automatically adapt to different environments to adjust audio parameters, dynamically suppress howling noise in real time, and improve spatial positioning and sound enhancement performance, thereby not only improving the quality and stability of the audio, but also enhancing the spatial effect of the audio and user experience.

[0004] To achieve the above object, the technical solution adopted by the present invention is: In one aspect, a digital audio processing system for audio includes an audio signal acquisition module, a signal analysis module, an adaptive adjustment module, a howling suppression module, and a sound effect optimization feedback module that are sequentially communicatively connected; The audio signal acquisition module is used to collect audio signal data of the sound system in real time through a sensor network, and the audio signal data includes spectrum, amplitude, phase, sound pressure level and reverberation time; The signal analysis module is used to perform spectrum feature decomposition on the audio signal data based on a non-negative matrix decomposition algorithm, and use a generative adversarial network algorithm to analyze the spectrum features, compare them with a standard spectrum curve, and generate a signal analysis report; The adaptive adjustment module is used to automatically adjust the gain, frequency response and equalization parameters of the audio signal using a fuzzy logic control algorithm based on the signal analysis report; The howling suppression module is used to detect the howling noise component in the audio signal in real time by using an adaptive feedback elimination algorithm, and to perform dynamic suppression; The sound optimization feedback module is used for a spatial sound enhancement algorithm based on high-order Bezier curve fitting, which optimizes the frequency, phase and spatial positioning of the audio signal by constructing a high-order Bezier curve model of multi-channel audio.

[0005] Furthermore, the operation process of the signal analysis module includes the following steps: Receiving the audio signal data and performing a preprocessing operation; The preprocessed audio signal data is decomposed into spectral features using the non-negative matrix decomposition algorithm to generate multiple independent spectral bases and corresponding weight matrices, extract spectral feature information of different frequency bands, and mark the main frequency components of the signal; Generative adversarial network algorithm is used to compare spectrum features with the standard spectrum curve built into the system in multiple layers. A high-precision feature matching model is generated through adversarial training to identify and highlight abnormal spectrum components, while generating stability and similarity scores for spectrum features. The analysis results are imported into the dynamic adjustment model, which outputs a feature difference graph based on the spectrum feature report generated by the generative adversarial network algorithm, and generates dynamic adjustment prompts based on the spectrum difference and stability score; The signal analysis report is generated by combining the comparison results with the dynamic adjustment detection information, including detailed spectral component change trends, abnormal frequency identification results, gain distribution suggestions, frequency response curve matching, and recommended adjustment ranges for each audio parameter.

[0006] Furthermore, the construction process of the non-negative matrix decomposition algorithm includes the following steps: The audio signal data is divided into a number of adaptive sub-bands by an intelligent frequency band division algorithm, and main spectrum features are extracted according to the characteristics of different frequency bands to form an initial feature matrix; The sparsity control mechanism is used to initialize the basis matrix, and the structure of the basis matrix is ​​adjusted through a unique sparse optimization function; In the decomposition iteration, the sparse constraint parameters and residual suppression weights are dynamically adjusted to control the noise impact during the decomposition process and gradually optimize the sparse structure of the spectrum basis matrix; The final basis matrix and weight matrix are integrated and output, and a complete set of spectral features is formed through multi-band feature aggregation, and differentiated feedback of spectral features is generated.

[0007] Furthermore, the formula of the fuzzy logic control algorithm is as follows: ; in, Indicates the optimal gain adjustment amount; and They represent the weights of spectrum characteristics and equalization characteristics in the overall adjustment respectively; n represents the number of fuzzy rules, corresponding to different frequency band characteristics; represents the weight of the i-th fuzzy rule; represents the spectral features calculated based on the input variable x; represents the spectrum feature weight function; and represents the integration interval; m represents the number of equilibrium features; Indicates that based on the input variables Calculated equilibrium characteristics.

[0008] Furthermore, the operation process of the howling suppression module includes the following steps: The howling component in the audio signal is detected in real time through an adaptive feedback elimination algorithm, the main frequency causing the howling is identified and located, and the feedback path model is constructed at the same time; Using the feedback path model, an adaptive filter is applied to generate a suppression signal that is in antiphase with the howling frequency, and the howling noise component is dynamically eliminated by adjusting the filter weight coefficient in real time; According to the change of howling frequency, the gain and frequency band of the filter are adaptively adjusted, and the feedback path model is continuously optimized through the adaptive feedback elimination algorithm to enhance the suppression effect; After feedback suppression is completed, a howling suppression report is generated, including the changing trend of the howling frequency, the feedback path stability score, the suppression effect of each frequency band, and optimization suggestions for the feedback path model.

[0009] Furthermore, the formula of the feedback path model is as follows: ; in, Indicates the output audio signal after howling is suppressed; Represents the original input audio signal; N represents the number of feedback paths; represents the adaptive gain control coefficient of the I-th feedback path; Indicates the degree of influence of different paths on the feedback signal; represents the delay time of the i-th feedback path; represents the suppression strength coefficient; Represents the amplitude power exponent of the feedback signal; Represents the feedback signal amplitude in the i-th feedback path.

[0010] Furthermore, the operation process of the sound effect optimization feedback module includes the following steps: The spectrum, phase and spatial positioning data of the audio are collected through multi-channel sensors, and a high-order Bezier curve fitting algorithm is used to generate a three-dimensional spatial sound model, converting the spatial distribution characteristics of the sound source into a multi-order control point matrix; Based on the control point matrix generated by modeling, the frequency response and phase of each frequency band are fine-tuned; Using multi-channel feedback signals, the Bezier curve control parameters are adaptively regulated to dynamically adjust the spatial distribution of the audio signal. Generate an optimization feedback report based on the results of sound optimization, including the gain distribution diagram of spatial sound enhancement, error analysis of each sound source positioning, optimized frequency response curve and recommended parameter adjustment range.

[0011] Furthermore, the formula of the three-dimensional spatial sound effect model is as follows: ; in, Represents the time t, frequency f and phase The positioning point of the audio signal in three-dimensional space; represents the i-th basis function of the Bezier curve; Represents the coordinate matrix of the i-th control point of the Bezier curve; represents the frequency parameter of the i-th frequency band; Represents the phase offset of the i-th frequency band; Indicates the attenuation coefficient and controls the distance Impact on audio signals; represents the control point at time t The dynamic distance function of the Bezier curve; t represents the time parameter; n represents the order of the Bezier curve.

[0012] On the other hand, a digital audio processing method for audio comprises the following steps: Collect audio signal data of speakers in real time through sensor networks; Decomposing the spectral features of the audio signal data based on a non-negative matrix decomposition algorithm, analyzing the spectral features using a generative adversarial network algorithm, comparing the spectral features with a standard spectral curve, and generating a signal analysis report; Based on the signal analysis report, a fuzzy logic control algorithm is used to automatically adjust the gain, frequency response and equalization parameters of the audio signal; Adopt adaptive feedback elimination algorithm to detect the howling noise component in the audio signal in real time and suppress it dynamically; The spatial sound enhancement algorithm based on high-order Bezier curve fitting optimizes the frequency, phase and spatial positioning of the audio signal by constructing a high-order Bezier curve model of multi-channel audio.

[0013] The beneficial effects of the present invention are as follows: the audio signal acquisition module of the present invention can collect multi-dimensional audio signal data such as spectrum, amplitude, phase, etc. in real time through a sensor network, ensure the comprehensiveness and accuracy of the data, and provide high-quality input for subsequent processing. The signal analysis module realizes spectrum feature decomposition through a non-negative matrix decomposition algorithm, and conducts in-depth analysis of spectrum features in combination with a generative adversarial network. This combined analysis method can more accurately identify the characteristics of audio signals, realize intelligent comparison with standard spectrum, quickly generate signal analysis reports, and improve the accuracy and efficiency of analysis. The adaptive adjustment module uses a fuzzy logic control algorithm to automatically adjust the gain, frequency response and equalization parameters of the audio. Through an intelligent adjustment mechanism, the system can adapt to different environments, optimize the output effect of the audio, and make the sound effect more consistent and improve the quality. The howling suppression module adopts an adaptive feedback elimination algorithm, which can quickly perform dynamic suppression when the howling noise is detected in real time, effectively reduce the howling interference in the audio, and enhance the user's auditory experience. The sound optimization feedback module uses a spatial sound enhancement algorithm based on high-order Bezier curve fitting to systematically build a spatial positioning model for multi-channel audio, making the frequency, phase and spatial positioning of the audio more precise and three-dimensional, significantly improving the spatial sense and immersive experience of the sound. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention is a module schematic diagram of a digital audio processing system for audio.

[0015] Figure 2 It is a flowchart of the operation process of the howling suppression module provided by an embodiment of the present invention.

[0016] Figure 3 The present invention is a flowchart of a digital audio processing method for audio. DETAILED DESCRIPTION

[0017] See also Figure 1-3 As shown, the present invention relates to a digital audio processing method and system for audio.

[0018] Example 1 A digital audio processing system for audio, comprising an audio signal acquisition module, a signal analysis module, an adaptive adjustment module, a howling suppression module and a sound effect optimization feedback module which are sequentially connected in communication; The audio signal acquisition module is used to collect audio signal data of the sound system in real time through a sensor network, and the audio signal data includes spectrum, amplitude, phase, sound pressure level and reverberation time; It should be noted that the audio signal data of the speaker is collected in real time through the sensor network, as follows: Sensor type: Various sensors are arranged in the sound system, including but not limited to microphones, pressure sensors, accelerometers, etc., to ensure that the spectrum, amplitude and other signal characteristics of the sound are captured from different angles.

[0019] Microphone array design: By configuring microphones of different numbers and positions to form a microphone array, the system can collect audio information from different directions and distinguish between direct sound and reflected sound to obtain high-quality environmental sound data.

[0020] Optimize the location of the sound source: Optimize the installation location of the sensor according to the environment in which the audio is used. For example, in a multi-channel system, sensors can be placed in multiple room corners to monitor the sound field conditions in different areas. This arrangement can improve the coverage and accuracy of audio collection.

[0021] Before the start of the acquisition, the sensor's acquisition environment is noise calibrated to establish a background noise baseline. Through real-time monitoring and comparison, the changes in environmental noise can be identified to improve the stability of signal acquisition. The sensor's sensitivity, frequency response and other parameters are regularly self-checked and calibrated to maintain the accuracy and stability of the sensor. For example, the frequency response of the sensor can be calibrated through an automated white noise test or a signal generator. The quality of the collected data and the signal strength are detected in real time through the feedback mechanism of the acquisition module. When the signal deviates from the preset range, the system automatically adjusts the acquisition parameters or issues a warning to ensure the accuracy and reliability of data acquisition.

[0022] The signal analysis module is used to perform spectrum feature decomposition on the audio signal data based on a non-negative matrix decomposition algorithm, and use a generative adversarial network algorithm to analyze the spectrum features, compare them with a standard spectrum curve, and generate a signal analysis report; The operation process of the signal analysis module includes the following steps: Receiving the audio signal data and performing a preprocessing operation; The preprocessed audio signal data is decomposed into spectral features using the non-negative matrix decomposition algorithm to generate multiple independent spectral bases and corresponding weight matrices, extract spectral feature information of different frequency bands, and mark the main frequency components of the signal; Generative adversarial network algorithm is used to compare spectrum features with the standard spectrum curve built into the system in multiple layers. A high-precision feature matching model is generated through adversarial training to identify and highlight abnormal spectrum components, while generating stability and similarity scores for spectrum features. Specifically, the generative adversarial network (GAN) algorithm is used, in which the generator is used to simulate the standard spectrum features, and the discriminator is used to compare the input spectrum features with the generated standard spectrum. The spectrum features obtained by NMF decomposition are compared with the standard spectrum curve built into the system at multiple levels. Through adversarial training between the generator and the discriminator, the GAN model gradually optimizes its feature recognition accuracy. The closer the features generated by the generator are to the standard spectrum curve, the easier it is for the discriminator to identify abnormal frequency components. Through multi-layer comparison, GAN can identify abnormal spectral components in the audio signal and highlight them. The GAN model generates a similarity score based on the degree of feature matching, and the stability score reflects the stability of the audio signal features over time. The scoring results are used to quantify the stability of the current signal features and the degree of matching with the standard curve.

[0023] The analysis results are imported into the dynamic adjustment model, which outputs a feature difference graph based on the spectrum feature report generated by the generative adversarial network algorithm, and generates dynamic adjustment prompts based on the spectrum difference and stability score; Specifically, the spectrum feature analysis results obtained by GAN comparison are input into the dynamic adjustment model for further processing. The dynamic adjustment model generates a feature difference map, which visually displays the difference between the current spectrum features and the standard spectrum to help identify the frequency bands that need to be adjusted. Based on the spectrum difference, stability score and similarity score, dynamic adjustment suggestions such as gain and equalization for specific frequency bands are generated. For example, if the high-frequency part of the spectrum deviates from the standard, the system will recommend reducing the high-frequency gain to ensure the quality of the sound output.

[0024] The signal analysis report is generated by combining the comparison results with the dynamic adjustment detection information, including detailed spectral component change trends, abnormal frequency identification results, gain distribution suggestions, frequency response curve matching, and recommended adjustment ranges for each audio parameter.

[0025] The construction process of the non-negative matrix decomposition algorithm includes the following steps: The audio signal data is divided into a number of adaptive sub-bands by an intelligent frequency band division algorithm, and main spectrum features are extracted according to the characteristics of different frequency bands to form an initial feature matrix; Specifically, according to the frequency range of the audio signal and the application requirements, the intelligent frequency band division algorithm is used to divide the audio signal into multiple adaptive sub-bands. For example, in the speech processing scenario, low-frequency (below 250 Hz), medium-frequency (250-4000 Hz) and high-frequency (above 4000 Hz) sub-bands are divided to adapt to the auditory sensitivity of the human ear. The main spectral features are extracted in each adaptive sub-band to form an initial feature matrix. This matrix is ​​used to describe the energy distribution of the audio signal in each frequency band.

[0026] The sparsity control mechanism is used to initialize the basis matrix, and the structure of the basis matrix is ​​adjusted through a unique sparse optimization function; Specifically, the basis matrix is ​​initialized through a sparsity control mechanism. The sparsity design of the basis matrix helps to extract the significant spectral features in the audio signal and reduce the interference of non-significant components. A dedicated sparse optimization function is designed to control the sparsity of the structure of the basis matrix so that the basis matrix can focus on reflecting the main frequency components in the signal.

[0027] In the decomposition iteration, the sparse constraint parameters and residual suppression weights are dynamically adjusted to control the noise impact during the decomposition process and gradually optimize the sparse structure of the spectrum basis matrix; Specifically, during the iterative process of non-negative matrix decomposition, the sparse constraint parameters are dynamically adjusted to ensure that the sparsity of the base matrix is ​​optimized at different iteration stages. According to the error in the decomposition result, the residual suppression weight is increased to control the influence of noise. For example, when the error is large, the residual suppression weight is increased to reduce the influence of noise on the base matrix. Through the dynamic adjustment of sparse constraints and residual suppression, the sparse structure of the base matrix is ​​gradually optimized, so that the final decomposed base matrix is ​​more representative and reflects the main characteristics of the audio signal.

[0028] The final basis matrix and weight matrix are integrated and output, and a complete set of spectral features is formed through multi-band feature aggregation, and differentiated feedback of spectral features is generated.

[0029] The adaptive adjustment module is used to automatically adjust the gain, frequency response and equalization parameters of the audio signal using a fuzzy logic control algorithm based on the signal analysis report; Specifically, according to the signal analysis report, the main parameters affecting the audio output are selected as the input of the fuzzy logic controller. These input parameters include spectrum deviation, signal amplitude, background noise level, etc.

[0030] Spectral Deviation: Describes the difference between the current audio signal and the standard spectrum, used to identify unbalanced frequency bands.

[0031] Signal Amplitude: It is used to judge whether the loudness of the audio signal is appropriate to avoid the volume being too loud or too soft.

[0032] Background noise level: By detecting the intensity of background noise, the controller can adjust the gain parameters to improve the sound quality in noisy environments.

[0033] Fuzzy rule base: Based on the input parameters, a series of fuzzy rules are designed. The fuzzy rule base includes some basic adjustment criteria, such as: If the spectrum deviation is large and the high frequency band is strong, reduce the high frequency gain.

[0034] If the signal amplitude is small and the low frequency band is weak, increase the low frequency gain.

[0035] If the background noise level is high and the signal amplitude is low, increase the gain appropriately to ensure output clarity.

[0036] Fuzzy reasoning mechanism: Utilize fuzzy reasoning method to generate output decision according to input fuzzy value and fuzzy rule base.

[0037] Furthermore, the formula of the fuzzy logic control algorithm is as follows: in, Indicates the optimal gain adjustment amount, which is used to automatically adjust the gain of the audio signal to ensure the best balance of audio effects; and They represent the weights of spectrum characteristics and equalization characteristics in the overall adjustment respectively; n represents the number of fuzzy rules, corresponding to different frequency band characteristics; represents the weight of the i-th fuzzy rule; represents the spectral features calculated based on the input variable x; represents the spectrum feature weight function; and represents the integral interval, the frequency range in which the i-th fuzzy rule applies; m represents the number of equilibrium features; Indicates that based on the input variables Calculated equilibrium characteristics.

[0038] The howling suppression module is used to detect the howling noise component in the audio signal in real time by using an adaptive feedback elimination algorithm, and to perform dynamic suppression; The operation process of the howling suppression module includes the following steps: The howling component in the audio signal is detected in real time through an adaptive feedback elimination algorithm, the main frequency causing the howling is identified and located, and the feedback path model is constructed at the same time; Specifically, the audio signal is monitored in real time through an adaptive feedback elimination algorithm to identify the howling noise component. Spectral analysis tools (such as fast Fourier transform) are used to decompose the audio signal and identify stable frequencies with significant peaks, which are usually the main components of howling. For the detected spectrum peaks, their frequency and amplitude characteristics are further analyzed to locate the main frequency that produces howling. Howling usually occurs in the higher frequency range, and the amplitude of this frequency component is more prominent. According to the howling frequency and its spectral characteristics, an adaptive feedback elimination algorithm is used to construct a feedback path model. The feedback path model is used to describe the acoustic feedback path between the microphone, speaker and environment, record the main frequency, gain, delay and other characteristics in the path, and provide a basis for subsequent howling suppression.

[0039] Using the feedback path model, an adaptive filter is applied to generate a suppression signal that is in antiphase with the howling frequency, and the howling noise component is dynamically eliminated by adjusting the filter weight coefficient in real time; Specifically, an adaptive filter is set to generate a suppression signal that is anti-phase with the detected howling frequency. The initial weight coefficient of the filter is determined by the parameters of the feedback path model, and the center frequency of the filter is set to the howling frequency in the initial stage. The suppression signal generated by the adaptive filter has the same characteristics as the howling frequency but opposite phase, which is used to offset the amplitude of the howling noise. This anti-phase suppression signal is consistent with the howling noise in frequency, but opposite in amplitude, thereby achieving an effective offset effect. The adaptive filter adjusts its weight coefficient in real time according to the audio signal to ensure that the frequency and amplitude of the anti-phase signal always match the howling frequency. This process can use an adaptive algorithm such as the minimum mean square error (LMS) algorithm or the RLS algorithm to ensure that the filter can quickly adjust its output signal when the howling frequency changes.

[0040] According to the change of howling frequency, the gain and frequency band of the filter are adaptively adjusted, and the feedback path model is continuously optimized through the adaptive feedback elimination algorithm to enhance the suppression effect; It should be noted that in the process of howling suppression, the change of howling frequency is identified by real-time monitoring of audio signals. When frequency drift is detected, the center frequency of the filter is adjusted to keep consistent with the new howling frequency. According to the intensity of the howling signal and the spectrum peak, the gain parameter of the filter is dynamically adjusted to ensure the stability of the suppression effect. When the howling intensity is large, the filter gain is increased to suppress the howling more strongly; when the howling signal weakens or disappears, the gain is appropriately reduced to reduce the impact on the normal audio signal. The frequency band range of the filter is continuously adjusted using the feedback path model, and the frequency band range is fine-tuned through the adaptive algorithm to avoid the filter suppressing too many frequency bands, thereby ensuring minimal interference to the normal audio signal. According to the change of the howling frequency and the adjustment process of the adaptive filter, the feedback path model is continuously optimized to record the latest howling frequency, gain, delay and other information. This dynamic feedback mechanism helps the system maintain an efficient suppression effect on howling in a constantly changing audio environment.

[0041] After feedback suppression is completed, a howling suppression report is generated, including the changing trend of the howling frequency, the feedback path stability score, the suppression effect of each frequency band, and optimization suggestions for the feedback path model.

[0042] Furthermore, the formula of the feedback path model is as follows: ; in, Indicates the output audio signal after howling is suppressed; Represents the original input audio signal; N represents the number of feedback paths; represents the adaptive gain control coefficient of the I-th feedback path, controlling the gain of howling suppression; Indicates the degree of influence of different paths on the feedback signal; represents the delay time of the i-th feedback path, that is, the delay of the signal on this path. The delay time is calculated by the signal analysis module according to the reverberation time of the sound environment to adapt to different spatial sound characteristics; Represents the inhibition strength coefficient, which controls the strength of feedback inhibition; Indicates the amplitude power exponent of the feedback signal, which is used for nonlinear adjustment of the feedback signal to achieve a stronger suppression effect on high-intensity howling feedback. Usually the value is 2 or 3, which is used to enhance the dynamic response of feedback elimination; Represents the feedback signal amplitude in the i-th feedback path.

[0043] in, The specific formula is as follows: in, represents the maximum gain control coefficient; represents the feedback sensitivity coefficient; represents the feedback signal amplitude of the i-th feedback path, and represents the howling component detected in the path.

[0044] The sound effect optimization feedback module is used for optimizing the frequency, phase and spatial positioning of the audio signal by constructing a high-order Bezier curve model of multi-channel audio based on a spatial sound effect enhancement algorithm of high-order Bezier curve fitting; The operation process of the sound effect optimization feedback module includes the following steps: The spectrum, phase and spatial positioning data of the audio are collected through multi-channel sensors, and a high-order Bezier curve fitting algorithm is used to generate a three-dimensional spatial sound model, converting the spatial distribution characteristics of the sound source into a multi-order control point matrix; It should be noted that the spatial positioning of each sound source is calculated using the phase difference and time difference collected by the multi-channel sensor. The spatial position of the sound source is refined through the triangulation positioning algorithm to obtain the coordinate information of each sound source in three-dimensional space. Based on the data collected by multiple channels, a three-dimensional spatial sound effect model is generated using a high-order Bezier curve fitting algorithm. In the specific implementation, the spatial distribution characteristics of the sound source are described by the multi-order control points of the Bezier curve, and the spatial positioning of the sound source is converted into a multi-order control point matrix. According to the frequency, phase and position characteristics of the sound source, a control point matrix of the Bezier curve is generated. The control point matrix describes the distribution and change trend of the sound source in space, providing a basis for subsequent frequency response and phase optimization.

[0045] Based on the control point matrix generated by modeling, the frequency response and phase of each frequency band are fine-tuned; Specifically, on the basis of the control point matrix, the frequency response of each frequency band is fine-tuned. By adjusting the position and weight of the control point, the frequency response of the signal in different frequency bands is changed to make it more in line with the goal of spatial sound enhancement. For example, the density of control points is increased in the low frequency band to enhance the immersion of low frequencies; the smoothness of control points is increased in the high frequency band to improve clarity. According to the control point matrix, the phase of each frequency band is gradually adjusted to make the spatial positioning of the sound source more accurate. Phase adjustment can be achieved by changing the phase distribution of control points in the Bezier curve to achieve the optimization effect of spatial sound effects. For example, in the overlapping area of ​​multiple sound sources, the sound source separation is achieved by adjusting the phase to avoid the degradation of sound quality caused by phase conflict. In the fine-tuning process, the overall consistency of the spatial sound effect is maintained to ensure that the audio signals of different frequency bands maintain a natural transition in frequency response and phase. By smoothing the distribution of control points, it is ensured that the sound effect model will not produce abrupt changes in sound quality due to fine-tuning.

[0046] Using multi-channel feedback signals, the Bezier curve control parameters are adaptively regulated to dynamically adjust the spatial distribution of the audio signal. Generate an optimization feedback report based on the results of sound optimization, including the gain distribution diagram of spatial sound enhancement, error analysis of each sound source positioning, optimized frequency response curve and recommended parameter adjustment range.

[0047] Furthermore, the formula of the three-dimensional spatial sound effect model is as follows: in, Represents the time t, frequency f and phase The positioning points of the audio signal in the three-dimensional space are used to describe the distribution of sound effects in different spatial positions, forming the effect of three-dimensional sound enhancement; Represents the i-th basis function of the Bezier curve, used to control different control points The weight of The coordinate matrix of the ith control point of the Bezier curve is used to define the three-dimensional position of the spatial sound enhancement to ensure that the spatial distribution characteristics of the audio signal match the control point; represents the frequency parameter of the i-th frequency band; Represents the phase offset of the i-th frequency band; Indicates the attenuation coefficient and controls the distance The effect on the audio signal is used to simulate the spatial attenuation effect of the sound source in the three-dimensional space model; represents the control point at time t The dynamic distance function, that is, the distance between the control point and the listening position, is calculated through the sound pressure level and reverberation time parameters; t represents the time parameter, ranging from [0, 1], which controls the curve generation process and is synchronized with the time characteristics of the audio signal; n represents the order of the Bezier curve.

[0048] Example 2 A digital audio processing method for audio, comprising the following steps: Collect audio signal data of speakers in real time through sensor networks; Decomposing the spectral features of the audio signal data based on a non-negative matrix decomposition algorithm, analyzing the spectral features using a generative adversarial network algorithm, comparing the spectral features with a standard spectral curve, and generating a signal analysis report; Based on the signal analysis report, a fuzzy logic control algorithm is used to automatically adjust the gain, frequency response and equalization parameters of the audio signal; Adopt adaptive feedback elimination algorithm to detect the howling noise component in the audio signal in real time and suppress it dynamically; The spatial sound enhancement algorithm based on high-order Bezier curve fitting optimizes the frequency, phase and spatial positioning of the audio signal by constructing a high-order Bezier curve model of multi-channel audio.

[0049] In this embodiment, a digital audio processing method for audio is applied to a digital audio processing system for audio described in Embodiment 1, which will not be described in detail here.

[0050] In summary, the audio signal acquisition module of the present invention can collect high-quality audio data from multiple directions and positions in real time through a multi-sensor array and optimized arrangement, covering different sound field areas, thereby improving the comprehensiveness and accuracy of the data. The non-negative matrix decomposition algorithm is used to decompose the spectral features, and the features are compared at multiple levels through a generative adversarial network to accurately identify and highlight abnormal spectral components. The generated spectral feature report provides a reliable basis for further dynamic adjustment of audio signals.

[0051] The adaptive adjustment module of the present invention adopts a fuzzy logic control algorithm to dynamically adjust the gain, frequency response and equalization parameters of the audio signal based on the analysis report, and automatically generates an optimization adjustment plan to ensure the balance, clarity and comfort of the sound output. The howling suppression module uses an adaptive feedback elimination algorithm to detect and suppress the howling noise components in real time, dynamically adjust the filter parameters, eliminate the howling components in the audio, and effectively enhance the stability and sound quality of the system. Through a high-order Bessel curve fitting algorithm, the sound optimization feedback module constructs a three-dimensional spatial sound model, dynamically optimizes the frequency response and phase of the audio signal, and realizes accurate positioning and enhancement of the sound source in three-dimensional space. The optimized spatial sound experience provides users with a more immersive auditory effect.

[0052] The system automatically calibrates sensors and audio signals, detects signal quality and provides feedback mechanisms to adjust acquisition and processing parameters in real time. Through analysis and optimization reports, the system can adaptively adjust key parameters to make the audio processing process more stable and intelligent.

[0053] The above implementation modes are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A digital audio processing system for audio, characterized in that: It includes an audio signal acquisition module, a signal analysis module, an adaptive adjustment module, a howling suppression module and a sound effect optimization feedback module which are sequentially connected in communication; The audio signal acquisition module is used to collect audio signal data of the sound system in real time through a sensor network, and the audio signal data includes spectrum, amplitude, phase, sound pressure level and reverberation time; The signal analysis module is used to perform spectrum feature decomposition on the audio signal data based on a non-negative matrix decomposition algorithm, and use a generative adversarial network algorithm to analyze the spectrum features, compare them with a standard spectrum curve, and generate a signal analysis report; The adaptive adjustment module is used to automatically adjust the gain, frequency response and equalization parameters of the audio signal using a fuzzy logic control algorithm based on the signal analysis report; The howling suppression module is used to detect the howling noise component in the audio signal in real time by using an adaptive feedback elimination algorithm, and to perform dynamic suppression; The sound optimization feedback module is used for a spatial sound enhancement algorithm based on high-order Bezier curve fitting, which optimizes the frequency, phase and spatial positioning of the audio signal by constructing a high-order Bezier curve model of multi-channel audio.

2. A digital audio processing system for audio according to claim 1, characterized in that: The operation process of the signal analysis module includes the following steps: Receiving the audio signal data and performing a preprocessing operation; The preprocessed audio signal data is decomposed into spectral features using the non-negative matrix decomposition algorithm to generate multiple independent spectral bases and corresponding weight matrices, extract spectral feature information of different frequency bands, and mark the main frequency components of the signal; Generative adversarial network algorithm is used to compare spectrum features with the standard spectrum curve built into the system in multiple layers. A high-precision feature matching model is generated through adversarial training to identify and highlight abnormal spectrum components, while generating stability and similarity scores for spectrum features. The analysis results are imported into the dynamic adjustment model, which outputs a feature difference graph based on the spectrum feature report generated by the generative adversarial network algorithm, and generates dynamic adjustment prompts based on the spectrum difference and stability score; The signal analysis report is generated by combining the comparison results with the dynamic adjustment detection information, including detailed spectral component change trends, abnormal frequency identification results, gain distribution suggestions, frequency response curve matching, and recommended adjustment ranges for each audio parameter.

3. A digital audio processing system for audio according to claim 2, characterized in that: The construction process of the non-negative matrix decomposition algorithm includes the following steps: The audio signal data is divided into a number of adaptive sub-bands by an intelligent frequency band division algorithm, and main spectrum features are extracted according to the characteristics of different frequency bands to form an initial feature matrix; The sparsity control mechanism is used to initialize the basis matrix, and the structure of the basis matrix is ​​adjusted through a unique sparse optimization function; In the decomposition iteration, the sparse constraint parameters and residual suppression weights are dynamically adjusted to control the noise impact during the decomposition process and gradually optimize the sparse structure of the spectrum basis matrix; The final basis matrix and weight matrix are integrated and output, and a complete set of spectral features is formed through multi-band feature aggregation, and differentiated feedback of spectral features is generated.

4. A digital audio processing system for audio according to claim 1, characterized in that: The formula of the fuzzy logic control algorithm is as follows: ； in, Indicates the optimal gain adjustment amount; and They represent the weights of spectrum characteristics and equalization characteristics in the overall adjustment respectively; n represents the number of fuzzy rules, corresponding to different frequency band characteristics; represents the weight of the i-th fuzzy rule; represents the spectral features calculated based on the input variable x; represents the spectrum feature weight function; and represents the integration interval; m represents the number of equilibrium features; Indicates that based on the input variables Calculated equilibrium characteristics.

5. A digital audio processing system for audio according to claim 1, characterized in that: The operation process of the howling suppression module includes the following steps: The howling component in the audio signal is detected in real time through an adaptive feedback elimination algorithm, the main frequency causing the howling is identified and located, and the feedback path model is constructed at the same time; Using the feedback path model, an adaptive filter is applied to generate a suppression signal that is in antiphase with the howling frequency, and the howling noise component is dynamically eliminated by adjusting the filter weight coefficient in real time; According to the change of howling frequency, the gain and frequency band of the filter are adaptively adjusted, and the feedback path model is continuously optimized through the adaptive feedback elimination algorithm to enhance the suppression effect; After feedback suppression is completed, a howling suppression report is generated, including the changing trend of the howling frequency, the feedback path stability score, the suppression effect of each frequency band, and optimization suggestions for the feedback path model.

6. A digital audio processing system for audio according to claim 5, characterized in that: The formula of the feedback path model is as follows: ； in, Indicates the output audio signal after howling is suppressed; Represents the input original audio signal; N represents the number of feedback paths; represents the adaptive gain control coefficient of the I-th feedback path; Indicates the degree of influence of different paths on the feedback signal; represents the delay time of the i-th feedback path; represents the suppression strength coefficient; Represents the amplitude power exponent of the feedback signal; Represents the feedback signal amplitude in the i-th feedback path.

7. A digital audio processing system for audio according to claim 1, characterized in that: The operation process of the sound effect optimization feedback module includes the following steps: The spectrum, phase and spatial positioning data of the audio are collected through multi-channel sensors, and a high-order Bezier curve fitting algorithm is used to generate a three-dimensional spatial sound model, converting the spatial distribution characteristics of the sound source into a multi-order control point matrix; Based on the control point matrix generated by modeling, the frequency response and phase of each frequency band are fine-tuned; Using multi-channel feedback signals, the Bezier curve control parameters are adaptively regulated to dynamically adjust the spatial distribution of the audio signal. Generate an optimization feedback report based on the results of sound optimization, including the gain distribution diagram of spatial sound enhancement, error analysis of each sound source positioning, optimized frequency response curve and recommended parameter adjustment range.

8. A digital audio processing system for audio according to claim 7, characterized in that: The formula of the three-dimensional spatial sound effect model is as follows: ； in, Represents the time t, frequency f and phase The positioning point of the audio signal in three-dimensional space; represents the i-th basis function of the Bezier curve; Represents the coordinate matrix of the i-th control point of the Bezier curve; represents the frequency parameter of the i-th frequency band; Represents the phase offset of the i-th frequency band; Indicates the attenuation coefficient and controls the distance Impact on audio signals; represents the control point at time t The dynamic distance function of the Bezier curve; t represents the time parameter; n represents the order of the Bezier curve.

9. A digital audio processing method for audio, characterized in that: The system is applied to a digital audio processing system for audio as claimed in any one of claims 1 to 8, comprising the following steps: Collect audio signal data of speakers in real time through sensor networks; Decomposing the spectral features of the audio signal data based on a non-negative matrix decomposition algorithm, analyzing the spectral features using a generative adversarial network algorithm, comparing the spectral features with a standard spectral curve, and generating a signal analysis report; Based on the signal analysis report, a fuzzy logic control algorithm is used to automatically adjust the gain, frequency response and equalization parameters of the audio signal; Adopt adaptive feedback elimination algorithm to detect the howling noise component in the audio signal in real time and suppress it dynamically; The spatial sound enhancement algorithm based on high-order Bezier curve fitting optimizes the frequency, phase and spatial positioning of the audio signal by constructing a high-order Bezier curve model of multi-channel audio.

Images