Headphone music compensation calibration method, calibration device and medium

Abstract

The present invention provides a headphone music compensation calibration method, a calibration device and a medium. The headphone music compensation calibration method can accurately identify the response characteristics of the headphone in different frequency bands by collecting and preprocessing the frequency response data of the headphone, then dividing the frequency bands and extracting key features. Combining the machine learning model and the physical model, the compensation amount required for each frequency band can be predicted and calculated, thereby generating compensation filter parameters adapted to different headphone characteristics. These parameters are stored and supported for dynamic calling, so that the headphones can call the preset parameters and compensation filter parameters during normal use, thereby ensuring the real-time application of the calibration effect and achieving stable consistency of sound quality and high-quality experience.

Description

Technical Field

[0001] The technical field of the present invention belongs to communication technology, and in particular, relates to a headphone music compensation calibration method, a calibration device and a medium. Background Art

[0002] In Bluetooth headphones with active noise reduction, when ANC (active noise reduction) is turned on, the FB (feedback) microphone not only picks up the noise in the ear canal, but also picks up the music signal played by the speaker. This causes the input signal of the feedback noise reduction path to contain both residual noise (i.e., effective signal) and music signal (i.e., invalid signal), making the noise reduction process complicated. Since the music signal picked up by the FB microphone is mistakenly treated as noise for processing, when the ANC function is turned on, the music signal actually heard by the human ear will change, especially in the low-frequency part. The low-frequency music signal will be reduced, and the amount of reduction will be consistent with the FB noise reduction amount, which will affect the sound quality of the music.

[0003] Most existing headphone ANC and MC (music compensation) solutions rely on preset parameters, without fully considering the differences in cavities and components during the headphone production process. This solution results in significant differences in the sound quality experience of different headphones before and after the ANC function is turned on and off, and the consistency of sound quality compensation is poor.

[0004] Existing methods mainly adjust headphones in laboratory environments, without optimizing for individual differences in headphones. This makes it difficult for the compensation effect to adapt to changes in different headphones or usage environments, resulting in unsatisfactory ANC and sound quality compensation effects in actual use. Summary of the invention

[0005] The object of the present invention is to provide a headphone music compensation calibration method, a calibration device and a medium, which can solve the above-mentioned problems.

[0006] In a first aspect of the present invention, a headphone music compensation calibration method is provided, the method comprising the following steps:

[0007] Collect the frequency response data of each headset and pre-process the collected frequency response data;

[0008] Divide the preprocessed frequency response data into frequency bands and extract the key features of each frequency band;

[0009] The machine learning model and the physical model are combined to predict the amount of compensation required for each headset in different frequency bands, and the compensation filter parameters that can adapt to different headsets in different frequency bands are calculated;

[0010] The compensation filter parameters are written into a memory and dynamic calling is supported.

[0011] Preferably, the headphone music compensation calibration method further includes the steps of:

[0012] Writing the preset parameters of the earphone into the fixed filter of the filter;

[0013] Writing the compensation filter parameters into the compensation filter of the filter;

[0014] When the earphone is in use, the preset parameters and the compensation filter parameters are called to adjust different frequency bands.

[0015] Preferably, in the headphone music compensation calibration method, the collected frequency response data includes at least one type of data including speaker frequency response data, SPK playback signal collected by a feedback noise reduction microphone, and feedback active noise reduction amount.

[0016] Preferably, in the headphone music compensation calibration method, the preprocessing includes denoising and smoothing the frequency response data; and the frequency band division of the preprocessed frequency response data includes segmenting the headphone frequency response data into low frequency, medium frequency and high frequency.

[0017] Preferably, in the headphone music compensation calibration method, the key characteristics of each frequency band include at least one of peak frequency, peak amplitude, frequency response curve and key frequency band transfer function.

[0018] Preferably, in the headphone music compensation calibration method, the combination of the machine learning model and the physical model to obtain the prediction results of the compensation amounts required for different frequency bands includes using the machine learning model to predict the corresponding first compensation value for each frequency band according to the key features of each headphone in different frequency bands;

[0019] Establishing a physical model for describing key characteristics of the earphone in different frequency bands, and fitting a corresponding second compensation value for each frequency band according to the physical model;

[0020] The prediction results of the first compensation value and the second compensation value are combined and calculated by at least one of a weighted average algorithm and a fusion algorithm to obtain compensation filter parameters that can adapt to different headphones in different frequency bands.

[0021] Preferably, in the headphone music compensation calibration method, the headphone music compensation calibration method uses speaker frequency response data or a feedback noise reduction microphone to collect an SPK playback signal as frequency response data, and the calibration method comprises the steps of:

[0022] The feedback noise reduction microphone collects the excitation signal sent by the test equipment to generate a first electrical signal Y1, and pre-processes the first electrical signal Y1 to reach the target state and saves it; when actually testing the product, the collected second electrical signal Y2 is recorded;

[0023] By comparing the first electrical signal Y1 and the second electrical signal Y2, a weighted compensation coefficient W1 is calculated, wherein the weighted compensation coefficient W1 includes a frequency-end weighted coefficient and a correlation coefficient between a test frequency response and a transfer function;

[0024] According to T1=(Y2 / Y1)W1, the compensation filter parameter T1 is solved; and the compensation filter parameter T1 is written into the memory.

[0025] Preferably, in the headphone music compensation calibration method, the headphone music compensation calibration method is for using the feedback active noise reduction amount as the frequency response data, and the calibration method comprises the steps of:

[0026] Calculate the feedback noise reduction amount P2 of the product to be calibrated and the feedback noise reduction amount P1 under the target state;

[0027] According to the change of noise reduction, the compensation filter parameter T2 is calculated using T2=(P2 / P1)W2, where W2 is the weighted compensation coefficient, including the weighting of the frequency band, the correlation coefficient between the noise reduction amount and the cavity change;

[0028] The compensation filter parameter T2 is written into the memory.

[0029] In a second aspect of the present invention, there is also provided a headphone compensation calibration device, comprising:

[0030] A collection unit, used to collect frequency response data of each headset;

[0031] A preprocessing unit, used for preprocessing the collected frequency response data;

[0032] A frequency band division unit is used to divide the frequency response data preprocessed by the preprocessing unit into frequency bands and extract key features of each frequency band;

[0033] A compensation prediction unit, which is used to combine the machine learning model and the physical model to obtain the prediction results of the compensation amount required for each headset in different frequency bands;

[0034] A compensation parameter calculation unit, used to calculate the compensation parameters of the filter of each earphone in different frequency bands according to the result of the compensation prediction unit;

[0035] A storage unit, used for storing the compensation parameters;

[0036] A control unit is used to support dynamic calling of the compensation parameters.

[0037] In a third aspect of the present invention, a computer-readable storage medium is provided, wherein program instructions are stored in the computer-readable storage medium, and when the program instructions are executed, they are used to implement the method of the first aspect.

[0038] Compared with the prior art, the present invention has at least the following technical effects:

[0039] By collecting and preprocessing the frequency response data of the headphones, dividing the frequency bands and extracting key features, the response characteristics of the headphones in different frequency bands can be accurately identified. Combining machine learning models and physical models, the amount of compensation required for each frequency band can be predicted and calculated, thereby generating compensation filter parameters that adapt to the characteristics of different headphones. These parameters are stored and supported for dynamic call, so that the headphones can call preset parameters and compensation filter parameters during normal use, thereby ensuring real-time application of the calibration effect and achieving stable and consistent sound quality and high-quality experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A method flow chart of a headphone music compensation calibration method according to an embodiment of the present invention;

[0041] Figure 2 A framework diagram of a currently commonly used headphone music compensation solution provided by an embodiment of the present invention;

[0042] Figure 3 A scheme framework diagram for performing calibration after the earphone cavity is assembled is provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following will be described in more detail with reference to the schematic diagram of a headphone music compensation calibration method, calibration device and medium of the present invention, wherein preferred embodiments of the present invention are shown, and it should be understood that those skilled in the art can modify the present invention described herein while still achieving the advantageous effects of the present invention. Therefore, the following description should be understood as being widely known to those skilled in the art, and not as a limitation of the present invention.

[0044] For the sake of clarity, not all features of the actual embodiments are described. In the following description, well-known functions and structures are not described in detail because they would clutter the invention with unnecessary detail. It should be recognized that in the development of any actual embodiment, a large number of implementation details must be made to achieve the developer's specific goals, such as changing from one embodiment to another according to the limitations of the relevant system or the relevant business. In addition, it should be recognized that such development work may be complex and time-consuming, but it is just a routine task for those skilled in the art.

[0045] The present invention is described in more detail in the following paragraphs by way of example with reference to the accompanying drawings. The advantages and features of the present invention will become more apparent from the following description. It should be noted that the accompanying drawings are in very simplified form and are not in exact proportions, and are only used to facilitate and clearly assist in illustrating the purpose of the embodiments of the present invention.

[0046] Embodiment 1

[0047] like Figure 1 As shown, in this embodiment 1, a headphone music compensation calibration method is proposed, and the method includes the following steps:

[0048] Collect the frequency response data of each headset and pre-process the collected frequency response data;

[0049] Divide the preprocessed frequency response data into frequency bands and extract the key features of each frequency band;

[0050] The machine learning model and the physical model are combined to predict the amount of compensation required for each headset in different frequency bands, and the compensation filter parameters that can adapt to different headsets in different frequency bands are calculated;

[0051] The compensation filter parameters are written into a memory and dynamic calling is supported.

[0052] Furthermore, the method also includes the steps of: writing the preset parameters of the earphone into the fixed filter of the filter; writing the compensation filter parameters into the compensation filter of the filter; when the earphone is in use, adjusting different frequency bands by calling the preset parameters and the compensation filter parameters.

[0053] Specifically, first, during the testing process of earphone production, the test equipment is used to collect at least one type of data in the frequency response data (also called frequency response data) of each earphone, and the frequency response data includes SPK-FR (speaker frequency response), FB-Mic (SPK playback signal collected by the feedback noise reduction microphone), and FB-ANC (feedback active noise reduction amount). Among them, SPK-FR is the frequency response characteristic of the earphone speaker, that is, the volume output of the earphone speaker at different frequencies; FB-Mic specifically refers to the frequency response characteristic of the speaker playback signal captured by the feedback noise reduction microphone in the earphone; FB-ANC specifically refers to the performance of the earphone feedback active noise reduction system, that is, the effect of the earphone in reducing external noise in noise reduction mode.

[0054] After that, a series of preprocessing operations are required for the collected raw frequency response data to improve the data quality. The preprocessing operations include: denoising and smoothing. Denoising can remove unnecessary noise in the data, which may come from the test environment or the equipment itself. In addition, by smoothing the data, abnormal fluctuations caused by measurement errors or other factors can be reduced.

[0055] It should be noted that if the test equipment is used to collect multiple types of frequency response data for each headset during the test, all the collected data must be normalized to conform to a unified standard dimension range to ensure that the data between different headsets can be compared and analyzed under the same standard, so as to facilitate the discovery of differences between headsets and make corresponding sound quality compensation. A series of preprocessing of the collected frequency response data can ensure that the collected headphone frequency response data is accurate and consistent, and can be used for subsequent sound quality compensation calculations and analysis. Generate an accurate sound quality feature profile for each headset, providing the necessary data support for personalized sound quality compensation.

[0056] In this embodiment, the preprocessed frequency response data is divided into frequency bands according to low frequency, medium frequency and high frequency. Specifically, since the compensation requirements in different frequency bands are different, in order to facilitate weighting of different frequency bands according to product characteristics, it is necessary to divide the preprocessed frequency response data into different frequency bands according to low frequency, medium frequency and high frequency. Weighted processing is performed on each frequency band according to the characteristics of the headphone product. The purpose of weighting is to give different levels of attention to each frequency band according to its importance in affecting the sound quality, and to adjust the noise reduction effect of different frequency bands, wherein the impact of the feedback noise reduction path on the music spectrum is frequency-dependent. According to acoustic principles and practical experience, the impact of the feedback noise reduction path on the low and medium frequency bands is more significant, while the impact on the high frequency band is relatively small. Weighted processing helps to more accurately adjust the sound quality of the headphones to meet specific sound quality goals. For example, the low frequency band may require more gain weights to compensate, while the high frequency band may require lower weights to reduce noise or distortion.

[0057] Afterwards, the data of each frequency band is analyzed to extract the key features in each frequency band. The key features of each frequency band include extracting at least one of the peak frequency, peak amplitude, frequency response curve and key frequency band transfer function. Among them, on the frequency response curve, the peak frequency refers to the frequency at which the response reaches the highest point, and the peak amplitude refers to the response size at this frequency point. Extracting the peak frequency and peak amplitude of each frequency band helps to identify the performance of the headphones at a specific frequency. The frequency response curve is used to show the volume output of the headphones at different frequencies. The key frequency band transfer function is used to describe how the headphones transmit sound signals within a specific frequency band, including information such as gain and phase, which is crucial for designing compensation filters.

[0058] Furthermore, a machine learning model is used to predict a corresponding first compensation value for each frequency band based on the key features of each headset in different frequency bands. Specifically, the machine learning model takes the frequency response characteristics of each headset as input data. The machine learning model predicts the first compensation value (also called the optimal compensation amount) required for the headset in different frequency bands based on the input frequency response characteristics. This means that the machine learning model can provide an optimized first compensation value for each frequency band based on the frequency response characteristics of the headset to improve the sound quality performance of the headset.

[0059] Furthermore, a physical model is established to describe the key characteristics of the headset in different frequency bands, and a corresponding second compensation value is fitted for each frequency band according to the physical model. Specifically, a physical model based on acoustic theory is established, and this physical model is used to describe the frequency response characteristics of the headset. According to the frequency response data such as SPK-FR, FB-Mic or FB-ANC, the second compensation values ​​of different frequency bands are calculated. Afterwards, an adaptive compensation algorithm is used to combine the predicted results of the first compensation value predicted by the machine learning model and the second compensation value fitted by the physical model, and a personalized compensation filter parameter is generated using a weighted average or fusion algorithm. The personalized compensation filter parameter can be adjusted according to the specific frequency response characteristics of each headset, thereby compensating for the sound quality changes caused by the differences in the cavity and components during the headset production process, so that the sound quality experience of different headsets before and after the ANC function is turned on and off is more consistent.

[0060] It should be noted that the compensation filter parameters are tailored for each headset and can cover multiple aspects such as low-frequency compensation, high-frequency smoothing, bandwidth adjustment, etc., to ensure that the headset can provide the best sound quality in the entire frequency band.

[0061] In a specific example, a headphone music compensation calibration method is proposed. The headphone music compensation calibration method only uses speaker frequency response data or feedback microphone frequency response data as frequency response data. The calibration method includes the following steps:

[0062] The feedback noise reduction microphone collects the excitation signal sent by the test equipment to generate a first electrical signal Y1, and pre-processes the first electrical signal Y1 to reach the target state and saves it; when actually testing the product, the collected second electrical signal Y2 is recorded;

[0063] By comparing the first electrical signal Y1 and the second electrical signal Y2, a weighted compensation coefficient W1 is calculated, wherein the weighted compensation coefficient W1 includes a frequency-end weighted coefficient and a correlation coefficient between a test frequency response and a transfer function;

[0064] According to T1=(Y2 / Y1)W1, the compensation filter parameter T1 is solved; and the compensation filter parameter T1 is written into the memory.

[0065] In this embodiment, if Figure 2 This is a commonly used headphone compensation scheme framework diagram, which depicts the key components and their interrelationships in the headphone music compensation process. The music compensation effect is mainly affected by changes in the headphone cavity, including:

[0066] Differences in headphone cavity consistency: The physical properties of different headphone cavities may affect the propagation of sound and the final auditory effect.

[0067] Consistency differences in SPK (speaker) units: Differences in the manufacturing and performance of speaker units will also affect the sound quality.

[0068] Consistency differences of FB_Mic (feedback microphone) units: The accuracy and performance of the feedback microphone also have an important impact on the music compensation effect.

[0069] Apart from Figure 2 In addition to the music compensation scheme of the headphones shown in the framework diagram of FIG, there is another scheme that realizes music compensation by calling different sound source EQ (equalizer) parameters in different modes. However, this scheme mainly involves calling different parameters, so it will not be described here.

[0070] Specifically, Figure 3 The diagram shows a scheme framework for calibration after the earphone cavity is assembled. First, an excitation signal is sent through the test equipment. This signal is a specific audio signal used to test the acoustic performance of the earphone cavity. Then, the sound signal emitted by the speaker (SPK) and passing through the earphone cavity is collected through the feedback microphone (FB_Mic). Before the test, it is necessary to pre-test and save the collected signal (recorded as the first electrical signal Y1) in the test equipment. It should be noted that the first electrical signal Y1 is the signal generated under the ideal state of the ideal product. When the product is actually tested, the new collected signal (recorded as the second electrical signal Y2) is recorded. After that, the required weighted compensation coefficient W1 is calculated by comparing the first electrical signal Y1 and the second electrical signal Y2. The weighted compensation coefficient W1 is used to adjust the second electrical signal Y2 to the target state of the first electrical signal Y1 to compensate for the influence of the earphone cavity on the sound. According to T1=(Y2 / Y1)W1, the compensation filter parameter T1 is solved. And the compensation filter parameter T1 is written into the memory.

[0071] It should be noted that this test method can be applied to the whole machine test and the semi-finished product cavity test. In the whole machine test, the excitation signal can be played through Bluetooth audio (BT Audio) or lead-in mode, or the built-in excitation signal of the headset can be played through instructions. It should be noted that when testing the whole machine, the ANC mode should be kept off to avoid the influence of the ANC system on the test results.

[0072] It should be noted that the test method can be further simplified by using a test microphone and a test fixture to directly collect the frequency response data of the speaker with the cavity. In this method, only the location of data collection changes, but the other test steps and calculation methods remain unchanged. In this case, the weighted compensation coefficient W1 includes the microphone-related response to ensure that the compensation filter can more accurately reflect the actual acoustic performance.

[0073] Figure 3 The framework diagram in the figure shows in a simplified way how the excitation signal is sent out through the speaker, collected by the feedback microphone, and how the acoustic performance is tested and compensated through the headphone cavity. This test solution helps to ensure that the headphone products maintain consistent sound quality performance in different production batches.

[0074] In another specific example, another headphone music compensation calibration method is proposed. The headphone music compensation calibration method is for using the feedback active noise reduction amount as the frequency response data. The calibration method includes the steps of:

[0075] Calculate the feedback noise reduction amount P2 of the product to be calibrated and the feedback noise reduction amount P1 under the target state;

[0076] According to the change of noise reduction, the compensation filter parameter T2 is calculated using T2=(P2 / P1)W2, where W2 is the weighted compensation coefficient, including the weighting of the frequency band, the correlation coefficient between the noise reduction amount and the cavity change;

[0077] The compensation filter parameter T2 is written into the memory.

[0078] Specifically, during the test, the headphones are placed in a dedicated test fixture and audio signals are played using external speakers. This method allows testers to evaluate the noise reduction performance of the headphones in a controlled environment.

[0079] like Figure 3 As shown, the earphones are placed in a test fixture. By playing an external sound source, the earphones switch between ANC_OFF (active noise reduction off) mode and FB_ANC (active noise reduction on) mode. Specifically, the signal PPNC is collected in the active noise reduction off (ANC_OFF) mode, and then the signal PFB_ANC is collected in the active noise reduction on (FB_ANC) mode.

[0080] By switching the working mode of the headset, the feedback active noise reduction amount P2 of the product to be calibrated and the feedback active noise reduction amount P1 in the target state can be obtained respectively. P1 is calculated by testing PPNC when active noise reduction is turned off and PFB_ANC when active noise reduction is turned on. P2 is calculated by testing PPNC when active noise reduction is turned off and PFB_ANC when active noise reduction is turned on. P2 calculated by these collected signals will be compared with P1 of the preset target noise reduction data, and the compensation filter parameter T2 is calculated by T2=(P2 / P1)W2. Among them, the weighted and compensation coefficient W2 includes the weighting of the frequency band, the correlation coefficient between the noise reduction amount and the cavity change. By comparing the FB_ANC noise reduction amount P2 of the product to be calibrated with the FB_ANC noise reduction amount P1 in the target state, the comprehensive difference after the coupling of the earphone cavity and the monomer difference can be indirectly calculated. Through this calibration scheme, it can ensure that the earphones have consistent noise reduction performance in different production batches, and improve the reliability and user experience of the product.

[0081] Embodiment 2

[0082] In the second embodiment, a headphone music compensation device is provided, which includes: a collection unit, a preprocessing unit, a frequency band division unit, a compensation prediction unit, a compensation parameter calculation unit, a storage unit and a control unit.

[0083] Specifically, the acquisition unit is used to collect the frequency response data of each headset. By collecting the frequency response data of each headset, it is possible to conveniently provide personalized calibration for each headset to compensate for its unique frequency response characteristics.

[0084] The preprocessing unit is used to preprocess the collected frequency response data. The preprocessing can remove environmental noise and abnormal fluctuations in the collection process of the frequency response data, making the collected frequency response data cleaner and improving the accuracy of subsequent analysis.

[0085] The frequency band division unit is used to divide the frequency response data preprocessed by the preprocessing unit into frequency bands and extract the key features of each frequency band. Since the compensation requirements in different frequency bands are different, different frequency bands need to be weighted according to product characteristics. Therefore, the frequency response data needs to be divided into different frequency bands for analysis.

[0086] The compensation prediction unit is used to combine the machine learning model and the physical model to predict the compensation amount required for each headset in different frequency bands. The specific method for predicting the required compensation amount can be referred to Example 1 and will not be described in detail here.

[0087] The compensation parameter calculation unit is used to calculate the compensation parameters of the filter of each earphone in different frequency bands according to the result of the compensation prediction unit. The specific method for calculating the compensation parameters can be referred to in Example 1, which will not be described in detail here.

[0088] A storage unit is used to store the compensation parameters.

[0089] A control unit is used to support dynamic calling of the compensation parameters.

[0090] The headphone music compensation device proposed in this embodiment can calculate the compensation parameters of the filter of each headphone in different frequency bands through the compensation parameter calculation unit, thereby realizing personalized calibration of different headphones, minimizing the difference in sound quality of the headphones in the ANC mode on and off, and ensuring that the end user has no significant difference in sound quality experience under different usage conditions.

[0091] Embodiment 3

[0092] The third embodiment of the present invention provides a computer-readable storage medium, which stores program instructions. When the program instructions are executed by a processor, some or all of the steps of the headphone music compensation calibration method in the first embodiment can be implemented, so that the headphones can call preset parameters and filter compensation parameters during normal use, thereby ensuring the real-time application of the calibration effect and achieving stable and consistent sound quality and high-quality experience.

[0093] In summary, the headphone music compensation calibration method, calibration device and medium provided in the present invention can accurately identify the response characteristics of the headphone in different frequency bands by collecting and preprocessing the frequency response data of the headphone, then dividing the frequency bands and extracting key features. Combining the machine learning model and the physical model, the compensation amount required for each frequency band can be predicted and calculated, thereby generating compensation filter parameters that adapt to the characteristics of different headphones. These parameters are stored and supported for dynamic calling, so that the headphones can call the preset parameters and filter compensation filter parameters during normal use, thereby ensuring the real-time application of the calibration effect and achieving stable and consistent sound quality and high-quality experience.

[0094] The above is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any technician in the relevant technical field, without departing from the scope of the technical solution of the present invention, makes any form of equivalent replacement or modification to the technical solution and technical content disclosed in the present invention, which does not depart from the content of the technical solution of the present invention and still falls within the protection scope of the present invention.

Claims

1. A headphone music compensation calibration method, characterized in that: The method comprises the following steps: Collect the frequency response data of each headset and pre-process the collected frequency response data; Divide the preprocessed frequency response data into frequency bands and extract the key features of each frequency band; The machine learning model and the physical model are combined to predict the amount of compensation required for each headset in different frequency bands, and the compensation filter parameters that can adapt to different headsets in different frequency bands are calculated; The compensation filter parameters are written into a memory and dynamic calling is supported.

2. The headphone music compensation calibration method according to claim 1, characterized in that: Also includes the steps: Writing the preset parameters of the earphone into the fixed filter of the filter; Writing the compensation filter parameters into the compensation filter of the filter; When the earphone is in use, the preset parameters and the compensation filter parameters are called to adjust different frequency bands.

3. The headphone music compensation calibration method according to claim 1, characterized in that: The collected frequency response data includes at least one type of data among the frequency response data of the loudspeaker, the SPK playback signal collected by the feedback noise reduction microphone, and the feedback active noise reduction amount.

4. The headphone music compensation calibration method according to claim 1, characterized in that: The preprocessing includes denoising and smoothing the frequency response data; The dividing the preprocessed frequency response data into frequency bands includes dividing the frequency response data of the earphone into low frequency, medium frequency and high frequency segments.

5. The headphone music compensation calibration method according to claim 1, characterized in that: The key feature of each frequency band includes at least one of a peak frequency, a peak amplitude, a frequency response curve and a key frequency band transfer function.

6. The headphone music compensation calibration method according to claim 1, characterized in that: The combining of the machine learning model and the physical model to obtain the prediction results of the compensation amounts required for different frequency bands includes using the machine learning model to predict the corresponding first compensation value for each frequency band according to the key features of each headset in different frequency bands; Establishing a physical model for describing key characteristics of the earphone in different frequency bands, and fitting a corresponding second compensation value for each frequency band according to the physical model; The prediction results of the first compensation value and the second compensation value are combined and calculated by at least one of a weighted average algorithm and a fusion algorithm to obtain compensation filter parameters that can adapt to different headphones in different frequency bands.

7. The headphone music compensation calibration method according to claim 1, characterized in that: The headphone music compensation calibration method adopts the speaker frequency response data or the feedback noise reduction microphone to collect the SPK playback signal as the frequency response data. The calibration method comprises the steps of: The feedback noise reduction microphone collects the excitation signal sent by the test device to generate a first electrical signal Y1, and pre-processes the first electrical signal Y1 to reach a target state and saves it; When actually testing the product, the collected second electrical signal Y2 is recorded; By comparing the first electrical signal Y1 and the second electrical signal Y2, a weighted compensation coefficient W1 is calculated, wherein the weighted compensation coefficient W1 includes a frequency-end weighted coefficient and a correlation coefficient between a test frequency response and a transfer function; According to T1=(Y2 / Y1)W1, the compensation filter parameter T1 is solved; and the compensation filter parameter T1 is written into the memory.

8. The headphone music compensation calibration method according to claim 1, characterized in that: The headphone music compensation calibration method adopts the feedback active noise reduction amount as frequency response data, and the calibration method comprises the steps of: Calculate the feedback noise reduction amount P2 of the product to be calibrated and the feedback noise reduction amount P1 under the target state; According to the change of noise reduction, the compensation filter parameter T2 is calculated using T2=(P2 / P1)W2, where W2 is the weighted compensation coefficient, including the weighting of the frequency band, the correlation coefficient between the noise reduction amount and the cavity change; The compensation filter parameter T2 is written into the memory.

9. A headphone compensation calibration device, characterized in that: include: A collection unit, used to collect frequency response data of each headset; A preprocessing unit, used for preprocessing the collected frequency response data; A frequency band division unit is used to divide the frequency response data preprocessed by the preprocessing unit into frequency bands and extract key features of each frequency band; A compensation prediction unit, which is used to combine the machine learning model and the physical model to obtain the prediction results of the compensation amount required for each headset in different frequency bands; A compensation parameter calculation unit, used to calculate compensation filter parameters that can adapt to different headphones in different frequency bands according to the result of the compensation prediction unit; A storage unit, used for storing the compensation filter parameters; A control unit is used to support dynamic calling of the compensation filter parameters.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to implement the method according to any one of claims 1 to 8.

Images