Sound effect adjusting method for high-performance audio loudspeaker

Abstract

The invention relates to the technical field of stereo devices, in particular to a sound effect adjusting method for a high-performance audio loudspeaker. The method comprises the following steps: firstly, inputting physical parameters of a loudspeaker unit, a cavity structure and a passive diaphragm, establishing a coupling acoustic model by adopting finite element simulation, and generating a theoretical frequency response curve; actual frequency response data of the sound box are acquired through analog-to-digital conversion and a sensor, and a core frequency band which has the greatest influence on sound quality is extracted by combining a Gaussian process regression method; extracting acoustic characteristic parameters based on a frequency response deviation function in the core frequency band, inputting the acoustic characteristic parameters to a PESQ model, and driving an LMS adaptive filtering algorithm to complete preliminary adjustment; and then carrying out secondary optimization on the full-band parameter space by adopting a particle swarm optimization algorithm, and finally obtaining an optimal sound effect adjustment result. The method has the advantages of being high in adjustment efficiency, high in parameter adaptive capacity, accurate in sound effect evaluation and the like, and is suitable for automatic sound effect optimization scenes of high-fidelity Bluetooth sound boxes, intelligent sound boxes and other products.

Description

Technical Field

[0001] The present invention relates to the technical field of stereo devices, and in particular to a sound effect adjustment method for high-performance audio speakers. Background Art

[0002] In the design process of modern audio products (such as Bluetooth speakers, smart speakers, and mobile devices), sound tuning, as a core component for enhancing the user's auditory experience, is gaining increasing attention. Traditional sound tuning methods rely heavily on manual experience for frequency response analysis and parameter adjustment. These methods suffer from long tuning cycles, a dependence on engineers' subjective judgment for tuning quality, and a lack of adaptability to actual acoustic structures. These methods struggle to meet the current market demand for high-fidelity, personalized sound output.

[0003] Furthermore, existing methods typically adjust uniformity parameters based on the average error across the entire frequency band, ignoring the critical role of certain frequency bands in the subjective perception of sound quality. This results in tuning results that, while improving overall performance, struggle to accurately optimize the frequency domain characteristics that users are most sensitive to. With the advancement of acoustic modeling, signal processing, and intelligent optimization algorithms, a systematic approach is needed that integrates acoustic simulation, actual response analysis, and data-driven tuning to achieve intelligent and refined adjustment of sound parameters, effectively improving the final listening quality and tuning efficiency of the sound system.

[0004] In order to solve the above problems, the present invention proposes a sound effect adjustment method for high-performance audio speakers. Summary of the Invention

[0005] This invention provides a sound tuning method for high-performance audio speakers. Finite element simulation is used to construct a theoretical frequency response model and collect actual frequency response data. Gaussian process regression is used to identify the core frequency band and extract key acoustic features. Initial tuning is then performed using the LMS adaptive filtering algorithm driven by the PESQ scoring model. Particle swarm optimization is then used to optimize parameters across the entire frequency band, achieving efficient and accurate sound tuning and improving sound quality.

[0006] To achieve the above object, the present invention provides the following technical solutions: A sound effect tuning method for a high-performance audio speaker, comprising: Input the speaker unit parameters, cavity structure parameters, and passive diaphragm component parameters, use finite element simulation to build an initial acoustic model, and generate a theoretical frequency response curve; Preset the initial gain parameters, frequency band equalization parameters and limiting parameters of the digital audio power amplifier module according to the sound source type and target sound effect characteristics; While playing the standard test audio signal, the analog-to-digital conversion unit collects the input audio signal, and the acceleration sensor collects the vibration response data of the passive diaphragm at each frequency point in real time to generate the actual frequency response curve; Comparing the actual frequency response curve with the theoretical frequency response curve, selecting the core frequency band that has the greatest impact on sound quality using the Gaussian process regression method, and calculating the acoustic characteristic parameters of the core frequency band based on the frequency response deviation function; The acoustic feature parameters are input into the PESQ scoring model. The sound effect evaluation score and parameter weights output by the model are used to drive the adaptive filtering algorithm to perform preliminary adjustments to the gain parameters, frequency band equalization parameters, and limiting parameters. After the adjustment results are scored by the sound quality evaluation model, the adjustment parameter space of the entire frequency band is re-adjusted based on the optimization objective function to obtain the optimal adjustment parameters.

[0007] Furthermore, the steps for constructing the initial acoustic model are: Input the physical parameters of the speaker unit, cavity structure and passive diaphragm assembly; Based on the input parameters, a coupled acoustic system model consisting of the speaker unit, cavity structure, and passive diaphragm is established using finite element analysis, and acoustic fluctuation boundary conditions are set based on the acoustic properties of the material and the vibration behavior of the diaphragm. Through simulation analysis, the frequency response of the system is calculated and the theoretical frequency response curve of the system is generated.

[0008] Furthermore, the core frequency band selection step is: Input the actual frequency response curve and the theoretical frequency response curve data, and establish the nonlinear relationship between the actual frequency response curve and the theoretical frequency response curve through the Gaussian process regression model; Based on the inference results of Gaussian process regression, the contribution of each frequency band to the sound quality is evaluated, and the frequency band with the greatest impact on the sound quality is selected as the core frequency band.

[0009] Furthermore, the calculation formula of the frequency response deviation function is: ; in, represents the frequency response deviation function, Indicates the The measured response amplitude at each frequency point is Indicates the The model predicts the response amplitude at each frequency point, Indicates the The weighting coefficient of each frequency point.

[0010] Furthermore, the steps of the preliminary adjustment are: Acquire acoustic feature parameters and compare them with the standard speech template library, and output a normalized scoring weight vector; The scoring weight vector is introduced into the LMS adaptive filtering algorithm, and the filter weight is updated according to the sound effect error and the gain parameter, frequency band equalization parameter and limiter parameter are dynamically adjusted. Stop the current iteration cycle based on the PESQ score change rate or weight convergence trend, and output the preliminary adjustment results.

[0011] Furthermore, the update formula of the filter weight is: ; in, Table No. The filter weight of the step, represents the basic step size factor, represents the scoring weight vector, Indicates the The output error of the step, Indicates the The input signal vector of the step.

[0012] Furthermore, the optimization objective function adopts the particle swarm optimization algorithm to perform parameter optimization, and the specific implementation steps are as follows: Set the optimization objective function, initialize the particle swarm, and set the digital audio amplifier module parameter set corresponding to each particle, including gain parameters, frequency band equalization parameters, and limiting parameters; According to the PESQ score and the output value of the deviation function, the fitness function value of each particle is calculated, and the individual optimal solution and the group optimal solution of the particle are updated; Iterate particle positions based on the velocity and position update formula to continuously adjust the control parameters of the audio amplifier module; When the set number of iterations is reached or the convergence condition is met, the optimal tuning parameters are output.

[0013] The beneficial effects of the present invention are: 1. This invention constructs a Gaussian process regression model to accurately describe the nonlinear mapping relationship between the actual and theoretical frequency response curves. Combined with Bayesian uncertainty inference, it can scientifically assess the contribution of each frequency band to sound quality, thereby selecting the core frequency bands most critical to sound perception. This selection mechanism significantly reduces redundant calculations, focusing on sound-sensitive frequency bands for optimization and tuning, improving tuning efficiency and targeting.

[0014] 2. For the core frequency band, this invention incorporates the LMS adaptive filtering algorithm for preliminary parameter tuning. It dynamically updates filter weights by combining acoustic feature parameters with a scoring weight vector, enabling precise adjustment of gain, equalization, and limiting parameters. This method rapidly converges to optimal sound output within a small number of iterations, improving the system's adaptability to environmental and speaker variations and laying the foundation for subsequent refined optimization across the entire frequency band.

[0015] 3. After completing preliminary tuning of the core frequency band, the present invention introduces a particle swarm optimization algorithm to globally optimize the parameter space across the entire frequency band. This algorithm uses PESQ scores and response deviations as fitness metrics and utilizes a particle position iteration strategy to optimize the multi-dimensional control parameters of the digital audio amplifier. This process effectively escapes local optima, resulting in comprehensive and high-quality tuning results, significantly improving final sound performance and user listening experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 This is a flow chart of a sound effect adjustment method for a high-performance audio speaker provided by the present invention. DETAILED DESCRIPTION

[0017] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0018] Example 1 A sound effect tuning method for high-performance audio speakers, such as Figure 1 Shown, including: S100: Input speaker unit parameters, cavity structure parameters, and passive diaphragm assembly parameters, use finite element simulation to build an initial acoustic model, and generate a theoretical frequency response curve; Furthermore, the steps for constructing the initial acoustic model are: Input the physical parameters of the speaker unit, cavity structure and passive diaphragm assembly; Based on the input parameters, a coupled acoustic system model consisting of the speaker unit, cavity structure, and passive diaphragm is established using finite element analysis, and acoustic fluctuation boundary conditions are set based on the acoustic properties of the material and the vibration behavior of the diaphragm. Through simulation analysis, the frequency response of the system is calculated and the theoretical frequency response curve of the system is generated.

[0019] Specifically, the input includes the following three types of key physical parameters: speaker unit parameters, such as diaphragm radius, magnetic circuit structure (magnet material, magnetic gap size), coil inductance / resistance, mechanical damping coefficient, effective mass, stiffness, etc.; cavity structure parameters, including cavity shape (cylindrical, square box, etc.), size (length, width and height), opening area, air hole position and number, structural wall thickness, material (such as ABS, PC, metal) and its elastic modulus and damping factor; passive diaphragm parameters, including diaphragm thickness, material properties (density, elastic modulus, Poisson's ratio), prestressed state, fixed boundary conditions (suspended edge / solid edge) and its coupling method with the cavity.

[0020] These parameters can be obtained through actual measurement, CAD modeling, standard device specifications or early system calibration.

[0021] Use commercial finite element simulation software (such as COMSOL Multiphysics, ANSYS, and ABAQUS) to build a three-dimensional structural model consisting of the speaker unit, cavity structure, and passive diaphragm. This model requires the following configuration: geometry modeling: import or draw the 3D geometry model to ensure that the dimensions and installation methods of each component are consistent; material assignment: assign acoustic and structural properties to each component based on the key physical parameters input above, including elastic modulus, density, Poisson's ratio, and structural damping; meshing: refine the model, especially using localized mesh refinement in vibration-concentrated areas (such as the center of the passive diaphragm) to improve simulation accuracy; and coupling boundary settings: set the structural-acoustic coupling surface between the speaker unit and the cavity, set the edges of the passive diaphragm as fixed or sliding boundary conditions, set the center as a free response area, define the speaker's sound-emitting surface as a structural driving source or velocity boundary, and set the cavity opening as a sound pressure release boundary or opening impedance boundary. A thermoacoustic module can also be introduced to consider the effects of temperature and humidity.

[0022] Set the analysis frequency range (e.g., 20 Hz to 20 kHz) and specify the frequency step. Apply a unit amplitude excitation input (e.g., sinusoidal vibration of a speaker diaphragm or a unit pressure source). Adjust parameters such as the iteration step size and convergence tolerance based on model complexity to ensure solution stability and accuracy. Enable the structure-acoustic coupling solver to obtain the cavity internal sound pressure distribution and diaphragm vibration response at each frequency point. After the simulation is complete, generate the system's theoretical frequency response curve.

[0023] Through the above modeling and simulation process, the multi-physics coupling behavior between the speaker, cavity, and passive diaphragm can be restored with high fidelity, and the theoretical frequency response results can be obtained in advance, avoiding a large amount of trial-and-error physical tuning process, and effectively improving the efficiency of sound tuning and the reliability of structural design.

[0024] S200: Presetting initial gain parameters, frequency band equalization parameters, and limiting parameters of the digital audio power amplifier module according to the sound source type and target sound effect characteristics.

[0025] Specifically, based on the target application scenario (such as voice calls, music playback, or movie sound effects) and the type of sound source used (such as human voice, electronic music, acoustic instruments, etc.), a sound effect characteristic requirement model is first established to clarify the enhancement or suppression characteristics required for each frequency band. Combining industry experience and historical adjustment data, the sound effect preference parameter template corresponding to the sound source type is extracted. Subsequently, based on the frequency response characteristics, cavity response, and electroacoustic characteristics of the speaker unit, the initial gain parameters (used to control the overall output intensity), frequency band equalization parameters (used to process low-frequency enhancement or mid-frequency recess), and limiting parameters (used to suppress signal overload) are determined.

[0026] S300: While playing a standard test audio signal, the analog-to-digital conversion unit collects the input audio signal. Simultaneously, the acceleration sensor collects the vibration response data of the passive diaphragm at various frequencies in real time to generate the actual frequency response curve. Specifically, white noise, swept frequency signals (linear or logarithmic sweep from 20Hz to 8kHz), or ITU standard speech test samples (such as P.50) are played to cover the target frequency response range and ensure the spectral energy balance of the excitation signal. A high-precision ADC (analog-to-digital conversion unit) with a sampling rate of no less than 44.1kHz and a resolution of 24 bits is used to acquire the digital audio signal at the speaker input end as the excitation signal reference. The acquired data is used for synchronous comparison with the actual response to ensure analysis consistency. A miniature accelerometer is installed at the center of the passive diaphragm to record the diaphragm vibration acceleration signal during playback. This signal is pre-amplified and input into the data acquisition system for synchronous sampling. The acquired acceleration signal is bandpass filtered (to remove background noise) and transformed with an FFT to extract the vibration response amplitude at each frequency point and generate the actual frequency response curve.

[0027] S400: comparing the actual frequency response curve with the theoretical frequency response curve, selecting a core frequency band that has the greatest impact on sound quality using a Gaussian process regression method, and calculating acoustic characteristic parameters of the core frequency band based on a frequency response deviation function; Furthermore, the core frequency band selection step is: Input the actual frequency response curve and the theoretical frequency response curve data, and establish the nonlinear relationship between the actual frequency response curve and the theoretical frequency response curve through the Gaussian process regression model; Based on the inference results of Gaussian process regression, the contribution of each frequency band to the sound quality is evaluated, and the frequency band with the greatest impact on the sound quality is selected as the core frequency band.

[0028] Specifically, the following two sets of data are collected: The actual frequency response curve data is the output response of the speaker system obtained through actual measurement, which is expressed as the vibration amplitude at different frequency points (such as 20Hz-20kHz); the theoretical frequency response curve data is the ideal response value obtained by the initial acoustic model simulation, which is consistent with the actual response point frequency, forming a comparable data pair. The two sets of data are recorded as and ,in, Indicates the frequency points.

[0029] Use the Gaussian process regression model to establish a nonlinear mapping relationship between theoretical response and actual response, including: setting the input variable Assume that the target output is the difference function ; Based on the training set , build the GPR model ,in is the mean function, is the kernel function (often using RBF kernel or Matern kernel), Represents any two frequency points; through the Bayesian inference mechanism, the error trend of all frequency points in the frequency domain is predicted, and the error confidence interval of each frequency point is obtained.

[0030] Based on the GPR inference results, the frequency response deviation and uncertainty index of each frequency point are calculated, including the frequency response deviation: ; Confidence weight ,in is the GPR prediction variance, and the comprehensive contribution score is defined as: , reflecting the weight of the frequency point's impact on the overall sound quality deviation. Set a contribution threshold (such as the top 10% cumulative contribution or a single point contribution greater than the set value) and select the continuous frequency band with the highest contribution score as the core frequency band.

[0031] By adopting the Gaussian process regression method, it is possible to accurately model the difference trend between theory and measurement under the conditions of nonlinear perturbations and noise in the frequency response data, thereby scientifically evaluating the contribution of each frequency band to the sound quality deviation, and then realizing a priority adjustment mechanism based on the core frequency band, avoiding repeated calculation of non-critical frequency bands.

[0032] Furthermore, the calculation formula of the frequency response deviation function is: ; in, represents the frequency response deviation function, Indicates the The measured response amplitude at each frequency point is Indicates the The model predicts the response amplitude at each frequency point, Indicates the The weighting coefficient of each frequency point.

[0033] Specifically, from the actual frequency response curve And theoretical frequency response curve Extract the frequency response amplitude data in the core frequency band, and record it as ,in , Indicates the core frequency band range. At each core frequency point Calculate the response deviation value ,in Can be obtained from PESQ model or empirical model.

[0034] Based on the deviation value set within the core frequency band , calculate the following representative acoustic characteristic parameters, including but not limited to: low-frequency response attenuation value, calculate the average or maximum value of the deviation value of the low-frequency area in the core frequency band; high-frequency resonance peak offset, identify the offset distance between the frequency response peak position and the theoretical peak; harmonic distortion rate estimation, detect the frequency response offset characteristics in the core frequency band, and estimate the total harmonic distortion trend.

[0035] By using the Gaussian process regression method to select the core frequency band that has the greatest impact on sound quality, and extracting representative acoustic feature parameters based on the frequency response deviation function, not only can the computational complexity be significantly reduced and the tuning efficiency be improved, but the parameter extraction results can also be ensured to be highly correlated with the subjective sound quality perception, which helps to achieve more accurate and efficient dynamic adjustment of sound effects.

[0036] S500: Input the acoustic feature parameters into the PESQ scoring model. The sound effect evaluation score and parameter weights output by the model are used to drive the adaptive filtering algorithm to perform preliminary adjustments to the gain parameters, frequency band equalization parameters, and limiting parameters. Furthermore, the steps of the preliminary adjustment are: Acquire acoustic feature parameters and compare them with the standard speech template library, and output a normalized scoring weight vector; The scoring weight vector is introduced into the LMS adaptive filtering algorithm, and the filter weight is updated according to the sound effect error and the gain parameter, frequency band equalization parameter and limiter parameter are dynamically adjusted. Stop the current iteration cycle based on the PESQ score change rate or weight convergence trend, and output the preliminary adjustment results.

[0037] Furthermore, the update formula of the filter weight is: ; in, Table No. The filter weight of the step, represents the basic step size factor, represents the scoring weight vector, Indicates the The output error of the step, Indicates the The input signal vector of the step.

[0038] Specifically, the PESQ scoring model is a subjective perception prediction model for speech quality defined in the ITU-T P.862 standard. Its core is to compare the perceptual differences between the reference signal and the signal to be evaluated, and output a score value of 0.5–4.5. In this implementation, the actual frequency response curve and the theoretical frequency response curve are constructed as input reference pairs, and acoustic feature parameters (low-frequency attenuation, high-frequency offset, harmonic distortion, etc.) are input into the model as perceptual offset indicators to simulate actual sound quality. The scoring model compares these parameters with the standard speech template library for perceptual differences, quantifies the contribution of each feature to subjective sound quality, and outputs a normalized parameter weight vector. The scoring weight vector is introduced into the LMS least mean square adaptive filtering algorithm to adjust the step size vector or gradient direction during the algorithm iteration process, namely: ; in, Table No. The filter weight of the step, represents the basic step size factor, represents the scoring weight vector, Indicates the The output error of the step, Indicates the The input signal vector of the step.

[0039] The system dynamically adjusts the above parameters in each iteration, aiming to minimize the error and maximize the score. The preliminary adjustment can be terminated when any of the following convergence criteria are met: The rate of change of PESQ score is lower than the set threshold (e.g. ); The parameter weight vector tends to be stable in several consecutive iterations (e.g. ).

[0040] The final output This is the preliminary tuning parameter set under the current sound conditions, providing a convergence initial value and directional reference for subsequent full-band optimization.

[0041] By introducing a joint mechanism of the PESQ scoring model and the LMS adaptive filtering algorithm, it is possible to achieve rapid and adaptive adjustment of sound effect parameters based on the acoustic characteristics of the core frequency band. This not only improves the tuning efficiency and reduces the waste of invalid computing resources, but also provides good initial parameters and directional guidance for subsequent full-band optimization.

[0042] S600: After the adjustment results are scored by the sound quality evaluation model, the adjustment parameter space of the entire frequency band is re-adjusted based on the optimization objective function to obtain the optimal adjustment parameters.

[0043] Furthermore, the optimization objective function adopts the particle swarm optimization algorithm to perform parameter optimization, and the specific implementation steps are as follows: Set the optimization objective function, initialize the particle swarm, and set the digital audio amplifier module parameter set corresponding to each particle, including gain parameters, frequency band equalization parameters, and limiting parameters; According to the PESQ score and the output value of the deviation function, the fitness function value of each particle is calculated, and the individual optimal solution and the group optimal solution of the particle are updated; Iterate particle positions based on the velocity and position update formula to continuously adjust the control parameters of the audio amplifier module; When the set number of iterations is reached or the convergence condition is met, the optimal tuning parameters are output.

[0044] Specifically, the optimization objective function Used to evaluate the parameter combination of power amplifier modules (in is the gain parameter, is the frequency band equalization parameter set, is the comprehensive impact of the limiting parameter set on the sound effect results, where the calculation formula of the objective function is: ; in, Indicates that the objective function is in the parameter combination The objective function value under Indicates the PESQ sound score, represents the output value of the deviation function, and Represents the weighting coefficient, and sets the particle swarm size to , each particle Corresponding to a set of control parameter vectors , and has an initial velocity and location , and randomly initialize the parameter values ​​of all particles within the set boundary range, for example: , , For each particle, use the current parameter combination to adjust the audio and play the test audio, and calculate the current fitness value based on the PESQ sound effect score and the output value of the deviation function If the current fitness is better than its historical best, then update its individual optimal position , if it is better than the global optimal solution, then update the group optimal position ; Use the standard PSO update formula to perform the next round of iteration on all particles: ; ; in, represents the inertia weight, and represents the learning factor, and Represents a random number in the interval [0,1]; when any of the following conditions is met, the iteration is terminated and the optimal parameter set is output: the set maximum number of iterations is reached , or the change in group fitness in the last five generations is lower than the threshold .

[0045] The full-band secondary tuning method based on the particle swarm optimization algorithm can further efficiently search for the optimal solution in the entire parameter space based on the preliminary tuning results, improve the parameter convergence speed and tuning accuracy, achieve global optimization of audio output quality, and effectively enhance the system's adaptability and sound quality performance in complex audio scenarios.

[0046] Example 2 During the development of a portable Bluetooth speaker designed for high-fidelity sound quality, the R&D team used the aforementioned sound tuning method to optimize the system sound. The specific implementation process is as follows: Based on the structural parameters of the speaker's φ52mm NdFeB speaker unit, ellipsoidal reflective cavity structure, and circular silicone passive diaphragm, the developers used the COMSOL Multiphysics finite element simulation platform to build an acoustic system model of the coupled speaker, cavity, and passive diaphragm. After setting the fluctuation boundary conditions for the material's sound velocity, density, and diaphragm tension, they ran a frequency-domain solver to calculate the system's theoretical frequency response curve over the 20Hz–20kHz range.

[0047] Based on the usage scenario where the portable speaker mainly plays vocals and pop music, the initial parameters of the preset digital audio amplifier module are: gain parameter: +6dB; frequency band equalization parameter: enhance the thickness of the vocals between 200 Hz and 400 Hz and reduce the harshness of high frequencies above 10 kHz; limiter parameter: peak value is limited to -2dBFS, and the release time is set to 20ms.

[0048] The speaker was placed in an anechoic chamber and played ITU standard speech and multi-frequency sweep signals. The audio input signal was acquired through an analog-to-digital conversion module. A MEMS triaxial accelerometer attached to the surface of the passive diaphragm measured the diaphragm's vibration response amplitude at each frequency point. A complete actual frequency response curve was generated after Fourier transform processing.

[0049] Using a Gaussian process regression model to compare the deviation between the actual frequency response and the theoretical response, significant deviations were found in the 160Hz–400Hz and 3.2kHz–6.5kHz ranges. Furthermore, acoustic characteristic parameters within the core frequency bands were calculated based on the frequency response deviation function, including a low-frequency response attenuation of -4.3dB, a high-frequency resonance shift of +280Hz, and a nonlinear harmonic distortion rate of 2.6%.

[0050] The above feature parameters were input into the pre-trained PESQ model, resulting in a current sound quality score of 3.26 (out of a maximum of 4.5). The model also output a normalized score weight vector. This weight vector was then introduced into the LMS adaptive filter. After 24 iterations, the PESQ score improved to 3.84. The system automatically determined convergence and output preliminary calibration results.

[0051] The tuned sound quality was re-evaluated, and an objective function was set. A particle swarm optimization algorithm was then used to optimize across the full-band parameter space. The particle swarm size was set to 40, and after 50 iterations, the optimal control parameters were output: the gain parameter was adjusted to +4.5dB; the band equalization parameters fine-tuned the low-frequency gain to +1.8dB, reducing the 4.8kHz peak; and the limiter delay was adjusted to 15ms, increasing output headroom by approximately 18%. The final PESQ score reached 4.21, reaching the industry standard for high-fidelity speakers.

[0052] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A sound effect tuning method for high-performance audio speakers, characterized in that: include: Input the speaker unit parameters, cavity structure parameters, and passive diaphragm assembly parameters, use finite element simulation to build an initial acoustic model, and generate a theoretical frequency response curve; Preset the initial gain parameters, frequency band equalization parameters and limiting parameters of the digital audio power amplifier module according to the sound source type and target sound effect characteristics; While playing the standard test audio signal, the analog-to-digital conversion unit collects the input audio signal, and the acceleration sensor collects the vibration response data of the passive diaphragm at each frequency point in real time to generate the actual frequency response curve; Comparing the actual frequency response curve with the theoretical frequency response curve, selecting the core frequency band that has the greatest impact on sound quality using the Gaussian process regression method, and calculating the acoustic characteristic parameters of the core frequency band based on the frequency response deviation function; The acoustic feature parameters are input into the PESQ scoring model. The sound effect evaluation score and parameter weights output by the model are used to drive the adaptive filtering algorithm to perform preliminary adjustments to the gain parameters, frequency band equalization parameters, and limiting parameters. After the adjustment results are scored by the sound quality evaluation model, the adjustment parameter space of the entire frequency band is re-adjusted based on the optimization objective function to obtain the optimal adjustment parameters.

2. The sound effect tuning method for a high-performance audio speaker according to claim 1, characterized in that: The steps for building the initial acoustic model are: Input the physical parameters of the speaker unit, cavity structure and passive diaphragm assembly; Based on the input parameters, a coupled acoustic system model consisting of the speaker unit, cavity structure, and passive diaphragm is established using finite element analysis, and acoustic fluctuation boundary conditions are set based on the acoustic properties of the material and the vibration behavior of the diaphragm. Through simulation analysis, the frequency response of the system is calculated and the theoretical frequency response curve of the system is generated.

3. The sound effect tuning method for a high-performance audio speaker according to claim 1, characterized in that: The steps for selecting the core frequency band are: Input the actual frequency response curve and the theoretical frequency response curve data, and establish the nonlinear relationship between the actual frequency response curve and the theoretical frequency response curve through the Gaussian process regression model; Based on the inference results of Gaussian process regression, the contribution of each frequency band to the sound quality is evaluated, and the frequency band with the greatest impact on the sound quality is selected as the core frequency band.

4. The sound effect tuning method for a high-performance audio speaker according to claim 1, characterized in that: The calculation formula of the frequency response deviation function is: ； in, represents the frequency response deviation function, Indicates the The measured response amplitude at each frequency point is Indicates the The model predicts the response amplitude at each frequency point, Indicates the The weighting coefficient of each frequency point.

5. The sound effect tuning method for high-performance audio speakers according to claim 1, characterized in that: The steps of the preliminary adjustment are: Acquire acoustic feature parameters and compare them with the standard speech template library, and output a normalized scoring weight vector; The scoring weight vector is introduced into the LMS adaptive filtering algorithm, and the filter weight is updated according to the sound effect error and the gain parameter, frequency band equalization parameter and limiter parameter are dynamically adjusted. Stop the current iteration cycle based on the PESQ score change rate or weight convergence trend, and output the preliminary adjustment results.

6. The sound effect tuning method for a high-performance audio speaker according to claim 5, characterized in that: The update formula for the filter weight is: ； in, Table No. The filter weight of the step, represents the basic step size factor, represents the scoring weight vector, Indicates the The output error of the step, Indicates the The input signal vector of the step.

7. The sound effect tuning method for a high-performance audio speaker according to claim 1, characterized in that: The optimization objective function adopts the particle swarm optimization algorithm to optimize the parameters, and the specific implementation steps are as follows: Set the optimization objective function, initialize the particle swarm, and set the digital audio amplifier module parameter set corresponding to each particle, including gain parameters, frequency band equalization parameters, and limiting parameters; According to the PESQ score and the output value of the deviation function, the fitness function value of each particle is calculated, and the individual optimal solution and the group optimal solution of the particle are updated; Iterate particle positions based on the velocity and position update formula to continuously adjust the control parameters of the audio amplifier module; When the set number of iterations is reached or the convergence condition is met, the optimal tuning parameters are output.

Images