A sound quality tuning system and method

Abstract

The application discloses a sound quality debugging system and method, and relates to the technical field of sound quality debugging.The method comprises the following steps: obtaining historical debugging records containing historical hardware feature vectors and their historical acoustic response curves; extracting a current hardware feature vector of a current sound; screening a plurality of historical debugging records based on the similarity with each historical hardware feature vector, and weighting and fusing the historical acoustic response curves to generate an initial acoustic response target curve; setting a debugging parameter, testing the current sound to obtain a current acoustic response curve, comparing the current acoustic response curve with the initial target curve to obtain an error vector; determining whether to output the debugging parameter or to iterate after optimizing and updating the parameter according to whether the error vector meets a preset condition; and the system comprises a historical debugging information acquisition module, a current feature extraction module, a target curve generation module, an error acquisition module and an optimization iteration module; and the application generates a personalized target curve through historical data matching and automatically iterates and optimizes, thereby improving the efficiency and accuracy of sound quality debugging.

Description

Technical Field

[0001] This invention relates to the field of sound quality adjustment technology, specifically a sound quality adjustment system and method. Background Technology

[0002] Sound quality tuning of audio equipment is a crucial step in ensuring that its final audio output meets design requirements. Traditional sound quality tuning heavily relies on the personal experience of the tuning engineer, adjusting various parameters such as gain at different frequencies and filter parameters through repeated listening and measurement. This method is not only inefficient, but also makes it difficult to guarantee consistent tuning results; different engineers or different batches of the same model of product may exhibit different sound quality performances.

[0003] With the development of audio processing technology, some automated sound quality tuning methods have emerged. These methods typically involve setting a fixed target response curve, then measuring the actual response of the current audio system and adjusting parameters to approximate that target curve. However, the limitation of this method lies in the fact that the target curve setting is often relatively simple and general, such as simply using standard curves like the Harman curve, failing to fully consider the impact of different audio hardware configurations (such as the type and number of different woofers and tweeters, cavity structure type, and audio channel configuration) on acoustic characteristics. Due to differences in hardware configuration, the same target curve may not achieve the ideal sound quality on different audio systems, and may even lead to difficulties in convergence during the tuning process or a poor final result.

[0004] Therefore, how to quickly generate personalized target curves that match the hardware characteristics of speakers with different hardware configurations, and how to optimize parameters efficiently and accurately, is a problem that needs to be solved in the field of sound quality tuning technology. Summary of the Invention

[0005] The purpose of this invention is to provide a sound quality adjustment system and method to solve the problems raised in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a sound quality tuning system and method. The methods include: Step S1: Obtain historical audio debugging information, which includes multiple historical debugging records. Each historical debugging record contains a historical hardware feature vector and a historical acoustic response curve corresponding to the historical hardware feature vector. The historical hardware feature vector is obtained by feature extraction from the hardware configuration information of the historical audio system. The hardware configuration information includes at least the type and number of woofers, the type and number of tweeters, the cavity structure type, and the audio channel configuration. Step S2: Obtain the hardware configuration information of the speaker to be debugged, and extract features from the hardware configuration information to generate the current hardware feature vector; Step S3: Calculate the similarity between the current hardware feature vector and each of the historical hardware feature vectors, and select the several historical debugging records with the highest similarity; use the similarity as the weight to perform weighted fusion on the several historical acoustic response curves to generate the initial acoustic response target curve of the current speaker to be debugged; Step S4: Set the initial value of the current sound quality adjustment parameter, apply the current sound quality adjustment parameter to the current speaker to be adjusted, obtain the current acoustic response curve by playing the test signal and collecting the actual acoustic response, and compare the current acoustic response curve with the initial acoustic response target curve to obtain the error vector; Step S5: Determine whether the preset conditions are met based on the error vector. If they are met, use the current sound quality adjustment parameters as the optimized sound quality adjustment parameters. If they are not met, update the current sound quality adjustment parameters using the optimization algorithm based on the error vector, and repeat steps S4 to S5 until the preset conditions are met.

[0007] Furthermore, step S1 includes: Step S1-1: Obtain the historical hardware configuration information of the audio system that has completed debugging, and discretize and encode each historical hardware feature in the historical hardware configuration information to generate the historical hardware feature vector V of the audio system that has completed debugging. hist =[v h1 ,v h2 ,...,v hn ], where n is the total number of historical hardware feature dimensions, v h1 ,v h2 ,...,v hn Each represents an encoded value for the historical hardware configuration information corresponding to n historical hardware feature dimensions. The historical hardware configuration information includes at least the type and number of woofers, the type and number of tweeters, the cavity structure type, and the audio channel configuration. Step S1-2: Play the test signal to the historically debugged audio system, collect the historical acoustic response of the historically debugged audio system, perform Fourier transform processing on the collected time-domain signal, and obtain the historical amplitude-frequency response curve H in the frequency domain. hist (f) represents the historical acoustic response curve of the historically tuned audio system after the tuning is completed, where f represents the frequency. Step S1-3: Transfer the historical hardware feature vector V hist With the historical acoustic response curve H hist (f) Associate storage to form a historical debugging record R hist ; Step S1-4: Repeat steps S1-1 to S1-3 to obtain several historical debugging records, which constitute the historical audio debugging information D={R hist,1 ,R hist,2 ,...,R hist,i ,...,R hist,m}, where R hist,1 ,R hist,2 ,...,R hist,i ,...,R hist,m These represent the 1st, 2nd, ..., ith, ..., mth historical debugging records, respectively.

[0008] Furthermore, step S2 includes: Step S2-1: Obtain the current hardware configuration information of the speaker to be tested. Following the same discretization encoding rules as in Step S1-1, encode each current hardware feature in the current hardware configuration information of the speaker to be tested to generate a current hardware feature vector V. cur =[v c1 ,v c2 ,...,v cn ], where n is the total number of current hardware feature dimensions, v c1 ,v c2 ,...,v cn These represent the encoded values ​​of the current hardware configuration information corresponding to the n current hardware feature dimensions, where each current hardware feature dimension v of the current hardware feature vector... c1 ,v c2 ,...,v cn The historical hardware feature dimensions v of the historical hardware feature vector h1 ,v h2 ,...,v hn They correspond in the same feature type order, i.e., v c1 With v h1 For the same type of hardware characteristics, v c2 With v h2 For the same type of hardware feature, and so on, v cn With v hn They correspond to the same type of hardware features, and the encoding rules are consistent across all corresponding dimensions.

[0009] Furthermore, step S3 includes: Step S3-1: For the i-th historical debugging record R hist,i The historical hardware feature vector is denoted as V. hist,i =[v h1,i ,v h2,i ,...,v hn,i], calculate the i-th historical debugging record R hist,i Historical hardware feature vector and current hardware feature vector V cur =[v c1 ,v c2 ,...,v cn The Euclidean distance d between them i The calculation formula is: And according to the Euclidean distance d i Calculate similarity S i The calculation formula is S i =1 / (1+d i ); Step S3-2: Based on the similarity S i The historical debugging records are sorted from high to low, and the K historical debugging records with the highest similarity are selected, where K is a preset positive integer; Step S3-3: For the selected K historical debugging records, calculate their similarity S. i As the weight, the historical acoustic response curves H corresponding to the K historical debugging records are... hist,i (f) Perform weighted fusion and generate the initial acoustic response target curve H of the speaker to be tuned according to the following formula. target (f): .

[0010] Furthermore, step S4 includes: Step S4-1: Set the initial values ​​of the current sound quality tuning parameters, which include the gain G(f) at each frequency point and the filter parameter Q; Step S4-2: Apply the current sound quality tuning parameters to the speaker to be tuned, play a test signal to the speaker, and collect the current actual acoustic response of the speaker when the test signal is played to obtain the current time-domain response signal s(t), where Indicates time; Step S4-3: Perform a Fourier transform on the current time-domain response signal s(t) to obtain the current amplitude-frequency response curve H in the frequency domain. meas (f) serves as the current acoustic response curve; Step S4-4: The current amplitude-frequency response curve H... meas (f) The initial acoustic response target curve H generated in step S3-3 target (f) Compare each frequency point, calculate the difference at each frequency point, and obtain the current error vector E(f)=H. meas (f)-H target (f).

[0011] Furthermore, step S5 includes: Step S5-1: Determine whether the error vector E(f) satisfies a preset condition, wherein the preset condition is that ∥E(f)∥<ε or the number of iterations reaches the upper limit N. max Where ||E(f)|| represents the norm of the error vector, ε is the preset threshold, and N max The maximum number of iterations is preset. If the preset conditions are met, the current sound quality tuning parameters are used as the optimized sound quality tuning parameters, and the process is terminated. If the preset conditions are not met, the sound quality tuning parameters are adjusted using the gradient descent method based on the current error vector E(f), specifically including: Step S5-2: For the gain G(f) at each frequency point, update it according to the following formula: G new (f)=G old (f)-α G ·E(f)·H meas (f), where α G The learning rate is the gain rate. Step S5-3: For the filter parameter Q, update it according to the following steps: update the current filter parameter Q. By increasing and decreasing a variable ΔQ used for numerical difference calculations, we obtain Q. old +ΔQ and Q old -ΔQ; This sets the two filter parameter values ​​Q to... old +ΔQ and Q old -ΔQ is applied to the current speaker to be debugged. Step S4 is repeated to measure the corresponding error norm. and Calculate the gradient approximation and update the parameters according to the following formula: , where α Q The filter learning rate; Step S5-4: Update the gain G of each frequency point new (f) and filter parameters Q new As the new current sound quality tuning parameters, steps S4 to S5 are repeated until the preset conditions are met, and finally the optimized sound quality tuning parameters are obtained.

[0012] The system includes: a historical debugging information acquisition module, a current feature extraction module, a target curve generation module, an error vector acquisition module, and an optimization iteration module; The historical debugging information acquisition module is used to acquire historical audio debugging information, which includes multiple historical debugging records. Each historical debugging record contains a historical hardware feature vector and a historical acoustic response curve corresponding to the historical hardware feature vector. The current feature extraction module is used to obtain the hardware configuration information of the speaker to be debugged and generate the current hardware feature vector; The target curve generation module is used to calculate the similarity between the current hardware feature vector and each historical hardware feature vector, filter several historical debugging records, and weight and fuse them to generate the initial acoustic response target curve. The error vector acquisition module is used to set and apply the current sound quality tuning parameters, measure the current acoustic response curve, and compare it with the initial target curve to obtain the error vector. The optimization iteration module is used to determine whether the preset conditions are met based on the error vector. If they are met, the optimized sound quality tuning parameters are output; otherwise, the parameters are updated and the tuning measurement module and optimization control module are repeatedly triggered until the conditions are met.

[0013] The historical debugging information acquisition module includes: Discretization coding unit, used to discretize and encode historical hardware configuration information to generate historical hardware feature vectors; The acoustic response acquisition unit is used to play the test signal and collect historical acoustic responses, and obtain the historical amplitude-frequency response curves through Fourier transform. The storage unit is used to associate historical hardware feature vectors with historical acoustic response curves and store them as historical debugging records. The record repeat unit is used to repeatedly execute the operation from obtaining historical hardware configuration information to associated storage, thereby obtaining multiple historical debugging records.

[0014] The current feature extraction module includes: The encoding unit is used to encode the current hardware configuration information according to the same discretization encoding rules as the historical debugging information acquisition module, and generate the current hardware feature vector. Each dimension of the current hardware feature vector corresponds to the dimensions of the historical hardware feature vector in the same feature type order.

[0015] The target curve generation module includes: The similarity calculation unit is used to calculate the distance between the historical hardware feature vector and the current hardware feature vector and determine the similarity. The filtering unit is used to sort historical debugging records by similarity and select a preset number of records. The weighted fusion unit is used to use the similarity of each selected historical debugging record as a weight to perform weighted fusion on each historical acoustic response curve to generate the initial acoustic response target curve.

[0016] The error acquisition module includes: The parameter setting unit is used to set the initial values ​​of the current sound quality tuning parameters, which include gain and filter parameters at each frequency point; The signal acquisition unit is used to play test signals and acquire the actual acoustic response of the current sound, thereby obtaining the current time-domain response signal. The transformation unit is used to perform a Fourier transform on the current time-domain response signal to obtain the current amplitude-frequency response curve; The comparison unit is used to subtract the current amplitude-frequency response curve from the initial acoustic response target curve at each frequency point, and arrange the differences at each frequency point in frequency order to generate an error vector.

[0017] The optimization and iteration module includes: The judgment unit is used to determine whether the error vector meets the preset conditions, including the error vector norm being less than a threshold or the number of iterations reaching the upper limit. The gain update unit is used to update the gain of each frequency point according to the error vector using the gradient descent method when the preset conditions are not met. The filter update unit is used to update the filter parameters using the numerical difference method when preset conditions are not met. The loop control unit is used to take the updated parameters as the new current sound quality tuning parameters and re-trigger the tuning measurement module and optimization control module.

[0018] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention constructs a database by extracting the hardware features (speaker configuration, cavity structure, etc.) of historical audio equipment and associating them with their acoustic response curves; it also extracts the same hardware features from new audio equipment, and generates personalized initial target curves by weighted fusion of historical curves based on similarity. This solves the problem of mismatch between fixed target curves and hardware, and improves the scientific nature of the debugging benchmark.

[0019] 2. After setting the initial parameters, the present invention obtains the current response curve by testing and compares it with the target curve to obtain the error vector. If the error does not meet the preset conditions, the gradient descent method is automatically used to update the frequency gain and the numerical difference method is used to update the filter parameters. The test and optimization are repeated until the target is met, realizing the adaptive adjustment of parameters and greatly improving the debugging efficiency and accuracy.

[0020] 3. This invention stores historical successful debugging records in a structured manner, filters similar cases through hardware feature matching and weighted fusion, so that new debugging can reuse historical experience; for speakers with the same hardware configuration, it can stably output consistent parameters, reduce reliance on human experience, and ensure the consistency of sound quality of different batches of products. Attached Figure Description

[0021] Fig. 1 This is a flowchart illustrating a sound quality adjustment method according to the present invention; Fig. 2 This is a schematic diagram of the structure of a sound quality adjustment system according to the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Example: Figs. 1-2 As shown, the present invention provides a technical solution, a sound quality adjustment system and method. The methods include: Step S1: Obtain historical audio debugging information, which includes multiple historical debugging records. Each historical debugging record contains a historical hardware feature vector and a historical acoustic response curve corresponding to the historical hardware feature vector. The historical hardware feature vector is obtained by feature extraction from the hardware configuration information of the historical audio system. The hardware configuration information includes at least the type and number of woofers, the type and number of tweeters, the cavity structure type, and the audio channel configuration. Step S1-1: Obtain the historical hardware configuration information of the speakers that have completed debugging. There are 3 historical debugging records. The hardware configuration information of each historical speaker is as follows: Historical debugging records R hist,1 Two paper cone woofers and one dome tweeter; sealed enclosure; 2.0 audio channels. Historical debugging records R hist,2 The woofer type is paper cone, quantity 1; the tweeter type is dome, quantity 1; the cavity structure type is bass-reflex; the audio channel configuration is 2.0; Historical debugging records R hist,3 The woofer is a metal cone type, with one unit; the tweeter is a horn type, with two units; the cavity structure is sealed; the audio channel configuration is 2.1. Each hardware feature is discretized and encoded according to the following rules: Subwoofer type: Paper cone = 01, Metal cone = 02; Number of subwoofers: 1 = 01, 2 = 02; Tweeter types: Dome = 01, Horn = 02; Number of tweeters: 1 = 01, 2 = 02; Cavity structure type: Closed type = 01, Phase-reversing type = 02; Audio channel configuration: 2.0=01, 2.1=02; Generate feature vectors for each historical hardware component: Vhist,1 =[01,02,01,01,01,01]; V hist,2 =[01,01,01,01,02,01]; V hist,3 =[02,01,02,02,01,02]; Step S1-2: Play the test signal (20Hz-20kHz sweep signal) to each of the previously debugged speakers, collect the historical acoustic response, and obtain the historical amplitude-frequency response curve through Fourier transform: H hist,1 (f): Due to the dual woofers, the low-frequency response is enhanced, resulting in a higher sound pressure level at 60Hz; H hist,2 (f): Due to the phase inversion structure, there is enhancement at 70Hz; H hist,3 (f): Due to the metal cone woofer and double horn tweeter, both low and high frequencies are improved; Steps S1-3: Associate and store each historical hardware feature vector with its corresponding historical acoustic response curve to form a historical debugging record R. hist,1 R hist,2 R hist,3 ; Steps S1-4: Obtain historical audio tuning information D={R hist,1 ,R hist,2 ,R hist,3}

[0024] Step S2: Obtain the hardware configuration information of the speaker to be debugged, and extract features from the hardware configuration information to generate the current hardware feature vector; Step S2-1: The hardware configuration information of the speaker to be tested is as follows: Subwoofer type: paper cone; Number of woofers: 1; Tweeter type: Dome; Number of tweeters: 1; Cavity structure type: Closed type; Audio channel configuration: 2.0; Encode according to the same encoding rules as in step S1-1 to generate the current hardware feature vector: V cur =[01,01,01,01,01,01]; where each dimension corresponds to the dimensions of the historical hardware feature vector in the same feature type order: v c1 =01 and v h1 All correspond to the type of subwoofer; v c2 =01 and vh2 Each corresponds to the number of subwoofers; v c3 =01 and v h3 All correspond to tweeter types; v c4 =01 and v h4 Each corresponds to the number of tweeters; v c5 =01 and v h5 All correspond to cavity structure types; v c6 =01 and v h6 All correspond to audio channel configurations.

[0025] Step S3: Calculate the similarity between the current hardware feature vector and each of the historical hardware feature vectors, and select the several historical debugging records with the highest similarity; use the similarity as the weight to perform weighted fusion on the several historical acoustic response curves to generate the initial acoustic response target curve of the current speaker to be debugged; Step S3-1: For each historical debugging record, calculate the Euclidean distance and similarity between its historical hardware feature vector and the current hardware feature vector; For R hist,1 V hist,1 =[01,02,01,01,01,01], and V cur The difference lies in the second dimension (number of woofers): S1 = 1 / (1+1) = 0.5; For R hist,2 V hist,2 =[01,01,01,01,02,01], and V cur The difference lies in the 5th dimension (cavity structure): S2 = 1 / (1+1) = 0.5; For R hist,3 V hist,3 = , with V cur The differences lie in the first dimension (woofer type), the third dimension (tweeter type), the fourth dimension (number of tweeters), and the sixth dimension (channel configuration): S3 = 1 / (1+2) ≈ 0.333; Step S3-2: Sort the records from highest to lowest similarity, select the K=2 historical debugging records with the highest similarity, and then select R. hist,1 (S1=0.5) and R hist,2 (S2=0.5); Step S3-3: Weighted fusion of the historical acoustic response curves from the two selected historical tuning records to generate the initial target acoustic response curve: ; At a frequency of 100Hz, H hist,1 (100) = 0.9 (relative sound pressure level); H hist,2 (100) = 0.8; Then: H target (100) = 0.5 × 0.9 + 0.5 × 0.8 = 0.45 + 0.4 = 0.85; Perform the same calculations at all frequency points to obtain the complete initial acoustic response target curve H. target (f).

[0026] Step S4: Set the initial value of the current sound quality adjustment parameter, apply the current sound quality adjustment parameter to the current speaker to be adjusted, obtain the current acoustic response curve by playing the test signal and collecting the actual acoustic response, and compare the current acoustic response curve with the initial acoustic response target curve to obtain the error vector; Step S4-1: Set the initial values ​​of the current sound quality tuning parameters; set the initial gain value of each frequency point G(f) = 1.0 (i.e., 0dB), and the filter parameter Q = 1.0; Step S4-2: Apply the current sound quality tuning parameters to the speaker to be tuned, play a 20Hz-20kHz sweep frequency signal to the speaker, and collect the actual acoustic response through the microphone to obtain the time domain response signal s(t); Step S4-3: Perform a Fourier transform on s(t) to obtain the current amplitude-frequency response curve H. meas (f); H was measured at 100 Hz meas (100) = 0.75; Step S4-4: Place H meas (f) and H target (f) Compare each frequency point and calculate the difference at each frequency point; at 100Hz: E(100)=H meas (100)-H target (100)=0.75-0.85=-0.1; Perform the same calculation on all frequency points to obtain the current error vector E(f).

[0027] Step S5: Determine whether the preset conditions are met based on the error vector. If they are met, use the current sound quality adjustment parameters as the optimized sound quality adjustment parameters. If they are not met, update the current sound quality adjustment parameters using the optimization algorithm based on the error vector, and repeat steps S4 to S5 until the preset conditions are met. Step S5-1: Determine if the error vector meets the preset conditions; calculate the norm of the error vector, assuming the preset threshold ε=0.1, the current ∥E(f)∥=0.2>0.1, and the current iteration count is 1, which has not reached the upper limit N. max =10, therefore the preset conditions are not met, and parameter update is required; Step S5-2: Update the gain G(f) at each frequency point using gradient descent; let the gain learning rate α be... G =0.1; at 100Hz: G new (100)=G old (100)-α G ·E(100)·H meas (100) = 1.0075; Step S5-3: Update the filter parameter Q using the numerical difference method; set the numerical difference change ΔQ = 0.05; change the current Q... old =1. Increase and decrease ΔQ respectively to obtain Q. old +ΔQ=1.05 and Q old -ΔQ=0.95; respectively Q old +ΔQ=1.05 and Q old -ΔQ=0.95 is applied to the current speaker to be debugged. Step S4 is executed again to measure the corresponding error norm; when Q old When +ΔQ=1.05, the measured value is... When Q old When -ΔQ=0.95, the measured value was... Calculate the gradient approximation and update the parameters; set the filter learning rate α. Q =0.1: ; Step S5-4: Update the gain G of each frequency point new (f) and filter parameters Q new =1.01 is used as the new current sound quality adjustment parameter, and steps S4 to S5 are repeated; after multiple iterations, when it is the 6th iteration, ∥E(f)∥=0.09<ε=0.1, which satisfies the preset condition. At this time, the optimized sound quality adjustment parameters are obtained: gain G at each frequency point. opt (f): 1.035 at 100Hz, with different optimized values ​​at other frequencies; filter parameter Q opt =1.03.

[0028] The system includes: a historical debugging information acquisition module, a current feature extraction module, a target curve generation module, an error vector acquisition module, and an optimization iteration module; The historical debugging information acquisition module is used to acquire historical audio debugging information, which includes multiple historical debugging records. Each historical debugging record contains a historical hardware feature vector and a historical acoustic response curve corresponding to the historical hardware feature vector. The current feature extraction module is used to obtain the hardware configuration information of the speaker to be debugged and generate the current hardware feature vector; The target curve generation module is used to calculate the similarity between the current hardware feature vector and each historical hardware feature vector, filter several historical debugging records, and weight and fuse them to generate the initial acoustic response target curve. The error vector acquisition module is used to set and apply the current sound quality tuning parameters, measure the current acoustic response curve, and compare it with the initial target curve to obtain the error vector. The optimization iteration module is used to determine whether the preset conditions are met based on the error vector. If they are met, the optimized sound quality tuning parameters are output; otherwise, the parameters are updated and the tuning measurement module and optimization control module are repeatedly triggered until the conditions are met.

[0029] The historical debugging information acquisition module includes: Discretization coding unit, used to discretize and encode historical hardware configuration information to generate historical hardware feature vectors; The acoustic response acquisition unit is used to play the test signal and collect historical acoustic responses, and obtain the historical amplitude-frequency response curves through Fourier transform. The storage unit is used to associate historical hardware feature vectors with historical acoustic response curves and store them as historical debugging records. The record repeat unit is used to repeatedly execute the operation from obtaining historical hardware configuration information to associated storage, thereby obtaining multiple historical debugging records.

[0030] The current feature extraction module includes: The encoding unit is used to encode the current hardware configuration information according to the same discretization encoding rules as the historical debugging information acquisition module, and generate the current hardware feature vector. Each dimension of the current hardware feature vector corresponds to the dimensions of the historical hardware feature vector in the same feature type order.

[0031] The target curve generation module includes: The similarity calculation unit is used to calculate the distance between the historical hardware feature vector and the current hardware feature vector and determine the similarity. The filtering unit is used to sort historical debugging records by similarity and select a preset number of records. The weighted fusion unit is used to use the similarity of each selected historical debugging record as a weight to perform weighted fusion on each historical acoustic response curve to generate the initial acoustic response target curve.

[0032] The error acquisition module includes: The parameter setting unit is used to set the initial values ​​of the current sound quality tuning parameters, which include gain and filter parameters at each frequency point; The signal acquisition unit is used to play test signals and acquire the actual acoustic response of the current sound, thereby obtaining the current time-domain response signal. The transformation unit is used to perform a Fourier transform on the current time-domain response signal to obtain the current amplitude-frequency response curve; The comparison unit is used to subtract the current amplitude-frequency response curve from the initial acoustic response target curve at each frequency point, and arrange the differences at each frequency point in frequency order to generate an error vector.

[0033] The optimization and iteration module includes: The judgment unit is used to determine whether the error vector meets the preset conditions, including the error vector norm being less than a threshold or the number of iterations reaching the upper limit. The gain update unit is used to update the gain of each frequency point according to the error vector using the gradient descent method when the preset conditions are not met. The filter update unit is used to update the filter parameters using the numerical difference method when preset conditions are not met. The loop control unit is used to take the updated parameters as the new current sound quality tuning parameters and re-trigger the tuning measurement module and optimization control module.

[0034] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for adjusting sound quality, characterized in that: Includes the following steps: Obtain historical audio tuning information. Several historical tuning records are available. Each historical tuning record contains a historical hardware feature vector and the corresponding historical acoustic response curve. Obtain the hardware configuration information of the speaker to be debugged and generate the current hardware feature vector; Calculate the similarity between the current hardware feature vector and each of the historical hardware feature vectors, sort them according to the similarity and select several historical debugging records, fuse the selected historical acoustic response curves, and generate the initial acoustic response target curve; Set the initial values ​​of the current sound quality tuning parameters and apply them to the current speaker. Measure the current acoustic response curve and compare it with the initial acoustic response target curve to obtain the error vector. When the preset conditions are met according to the error vector, the current sound quality adjustment parameters are used as the optimized sound quality adjustment parameters; otherwise, the current sound quality adjustment parameters are updated according to the error vector, and the measurement and update steps are repeated until the preset conditions are met.

2. The sound quality adjustment method according to claim 1, characterized in that: The acquisition of historical audio tuning information includes: The system retrieves historical hardware configuration information of audio equipment that has already undergone debugging, and discretizes and encodes various historical hardware features to generate historical hardware feature vectors. It plays test signals to collect historical acoustic responses, and obtains historical amplitude-frequency response curves in the frequency domain through Fourier transform as historical acoustic response curves. It associates and stores historical hardware feature vectors with historical acoustic response curves to form a historical debugging record. The system repeats the steps of retrieving historical hardware configuration information and storing it in the associated database to obtain several historical debugging records.

3. The sound quality adjustment method according to claim 2, characterized in that: The generation of the current hardware feature vector includes: Obtain the hardware configuration information of the speaker to be debugged, encode each hardware feature according to the same discretization coding rule as in claim 2, and generate the current hardware feature vector, wherein each dimension of the current hardware feature vector corresponds to each dimension of the historical hardware feature vector in the same feature type order.

4. The sound quality adjustment method according to claim 1, characterized in that: The calculation of similarity, sorting and selecting records, and generating the target curve include: Calculate the distance between the historical hardware feature vector of each historical debugging record and the current hardware feature vector, and determine the similarity based on the distance; sort the historical debugging records from high to low similarity, and select a preset number of historical debugging records; use the similarity of each selected historical debugging record as a weight to perform weighted fusion of each historical acoustic response curve to generate an initial acoustic response target curve.

5. The sound quality adjustment method according to claim 1, characterized in that: The obtained error vector includes: Set the initial values ​​of the current sound quality tuning parameters, which include gain and filter parameters at each frequency point; apply the current sound quality tuning parameters to the current speaker to be tuned, play the test signal and collect the actual acoustic response to obtain the current time domain response signal; perform a Fourier transform on the current time domain response signal to obtain the current amplitude-frequency response curve as the current acoustic response curve; subtract the current amplitude-frequency response curve from the initial acoustic response target curve at each frequency point, arrange the differences at each frequency point in frequency order, and generate an error vector.

6. The sound quality adjustment method according to claim 1, characterized in that: The step of updating the current sound quality adjustment parameters based on the error vector includes: Determine whether the error vector meets preset conditions, including the norm of the error vector being less than a threshold or the number of iterations reaching an upper limit; if not, update the sound quality tuning parameters using the gradient descent method based on the current error vector, including updating the gain of each frequency point and updating the filter parameters using the numerical difference method; repeat the process using the updated parameters as the current sound quality tuning parameters.

7. A sound quality adjustment system for performing a sound quality adjustment method according to any one of claims 1-6, characterized in that: The system includes: The system includes a historical debugging information acquisition module, a current feature extraction module, a target curve generation module, an error vector acquisition module, and an optimization iteration module. The historical debugging information acquisition module is used to acquire historical audio debugging information, which includes multiple historical debugging records. Each historical debugging record contains a historical hardware feature vector and a historical acoustic response curve corresponding to the historical hardware feature vector. The current feature extraction module is used to obtain the hardware configuration information of the speaker to be debugged and generate the current hardware feature vector; The target curve generation module is used to calculate the similarity between the current hardware feature vector and each historical hardware feature vector, filter several historical debugging records, and weight and fuse them to generate the initial acoustic response target curve. The error vector acquisition module is used to set and apply the current sound quality tuning parameters, measure the current acoustic response curve, and compare it with the initial target curve to obtain the error vector. The optimization iteration module is used to determine whether the preset conditions are met based on the error vector. If they are met, the optimized sound quality tuning parameters are output; otherwise, the parameters are updated and the tuning measurement module and optimization control module are repeatedly triggered until the conditions are met.

8. The sound quality adjustment system according to claim 7, characterized in that: The historical debugging information acquisition module includes: Discretization coding unit, used to discretize and encode historical hardware configuration information to generate historical hardware feature vectors; The acoustic response acquisition unit is used to play the test signal and collect historical acoustic responses, and obtain the historical amplitude-frequency response curves through Fourier transform. The storage unit is used to associate historical hardware feature vectors with historical acoustic response curves and store them as historical debugging records. The record repeating unit is used to repeatedly execute the operation from obtaining historical hardware configuration information to associated storage, thereby obtaining multiple historical debugging records; The current feature extraction module includes: The encoding unit is used to encode the current hardware configuration information according to the same discretization encoding rules as the historical debugging information acquisition module, and generate the current hardware feature vector. Each dimension of the current hardware feature vector corresponds to the dimensions of the historical hardware feature vector in the same feature type order.

9. The sound quality adjustment system according to claim 7, characterized in that: The target curve generation module includes: The similarity calculation unit is used to calculate the distance between the historical hardware feature vector and the current hardware feature vector and determine the similarity. The filtering unit is used to sort historical debugging records by similarity and select a preset number of records. The weighted fusion unit is used to use the similarity of each selected historical debugging record as a weight to perform weighted fusion on each historical acoustic response curve to generate the initial acoustic response target curve. The error acquisition module includes: The parameter setting unit is used to set the initial values ​​of the current sound quality tuning parameters, which include gain and filter parameters at each frequency point; The signal acquisition unit is used to play test signals and acquire the actual acoustic response of the current sound, thereby obtaining the current time-domain response signal. The transformation unit is used to perform a Fourier transform on the current time-domain response signal to obtain the current amplitude-frequency response curve; The comparison unit is used to subtract the current amplitude-frequency response curve from the initial acoustic response target curve at each frequency point, and arrange the differences at each frequency point in frequency order to generate an error vector.

10. A sound quality adjustment system according to claim 7, characterized in that: The optimization iteration module includes: The judgment unit is used to determine whether the error vector meets the preset conditions, including the error vector norm being less than a threshold or the number of iterations reaching the upper limit. The gain update unit is used to update the gain of each frequency point according to the error vector using the gradient descent method when the preset conditions are not met. The filter update unit is used to update the filter parameters using the numerical difference method when preset conditions are not met. The loop control unit is used to take the updated parameters as the new current sound quality tuning parameters and re-trigger the tuning measurement module and optimization control module.

Images