Optimized audio acquisition method and device based on Bluetooth headset, equipment and medium

Abstract

The invention relates to the technical field of earphones and the technical field of artificial intelligence, and discloses an optimized audio acquisition method and device based on a Bluetooth earphone, equipment and a medium, and the method comprises the steps: obtaining a real optimized audio corresponding to a first music audio; obtaining a prediction optimization audio output by the deep learning model based on the first music audio; obtaining a loss value between the predicted optimized audio and the real optimized audio; when the loss value meets a preset condition, stopping adjusting the model parameters of the deep learning model to obtain adjusted model parameters, and selecting the deep learning model loaded with the adjusted model parameters as an audio optimization model; and connecting the Bluetooth earphone, obtaining a second music audio to be played, when the music type identifier of the second music audio is the classical music identifier, obtaining a current optimized audio output by the audio optimization model based on the second music audio, and sending the current optimized audio to the Bluetooth earphone. According to the invention, the acquisition efficiency of the current optimized audio is improved.

Description

Technical Field

[0001] The present invention relates to the fields of earphone technology and artificial intelligence technology, and in particular to a method, device, equipment and medium for optimizing audio acquisition based on a Bluetooth earphone. Background Art

[0002] Optimized audio refers to audio files that have been processed by specific technical means to improve their sound quality. Bluetooth headphones are an audio device that uses Bluetooth technology to achieve wireless connection. The portability and sound insulation of Bluetooth headphones allow the delicate melody of optimized audio to be perfectly presented, creating an immersive music experience space for users.

[0003] However, the current process of obtaining optimized audio is cumbersome, which is not conducive to improving the efficiency of obtaining optimized audio.

[0004] The reason is that the current optimized audio acquisition process relies on manual operation, which consumes a lot of human resources and time resources, and increases the acquisition time of the current optimized audio. Therefore, it is not conducive to improving the acquisition efficiency of the current optimized audio, nor is it conducive to improving the sending efficiency of the current optimized audio. Summary of the invention

[0005] The present invention provides a method, device, computer equipment and storage medium for obtaining optimized audio based on a Bluetooth headset, so as to solve the technical problem that the current optimized audio obtaining process is cumbersome and is not conducive to improving the efficiency of obtaining the optimized audio.

[0006] In a first aspect, a method for optimizing audio acquisition based on a Bluetooth headset is provided, comprising: Acquire a first music audio, and when the music type identifier of the first music audio is a classical music identifier, acquire a real optimized audio corresponding to the first music audio; Inputting the first music audio into a deep learning model, and obtaining a predicted optimized audio output by the deep learning model based on the first music audio; Obtaining a loss value between the predicted optimized audio and the real optimized audio; Based on the loss value, adjusting model parameters of the deep learning model; When the loss value meets a preset condition, stop adjusting the model parameters of the deep learning model, obtain the adjusted model parameters, and select the deep learning model loaded with the adjusted model parameters as the audio optimization model; Connect a Bluetooth headset and obtain a second music audio to be played. When the music type identifier of the second music audio is a classical music identifier, obtain the current optimized audio output by the audio optimization model based on the second music audio, and send the current optimized audio to the Bluetooth headset.

[0007] Exemplarily, the acquiring of the first music audio, when the music type identifier of the first music audio is a classical music identifier, acquiring the real optimized audio corresponding to the first music audio includes: Acquire a first music audio, and when the music type identifier of the first music audio is a classical music identifier, acquire a plurality of test audios corresponding to the first music audio; Obtain the current storage space, current rating, current playback frequency, and current download count of the test audio; Adding the ratio of the current storage space to the preset storage space, the ratio of the current score to the preset score, the ratio of the current playback frequency to the current playback frequency, and the ratio of the current number of downloads to the preset number of downloads to generate a recommendation coefficient corresponding to the test audio; Obtain the recommendation coefficient corresponding to each of the test audios, and select the test audio with the largest recommendation coefficient as the true optimized audio corresponding to the first music audio.

[0008] Furthermore, the step of inputting the first music audio into a deep learning model and obtaining a predicted optimized audio output by the deep learning model based on the first music audio includes: Performing audio feature extraction on the first music audio to obtain a first feature vector; The first feature vector is input into a deep learning model, and the first feature vector is processed by the deep learning model to obtain predicted optimized audio.

[0009] Further, the obtaining of the loss value between the predicted optimized audio and the real optimized audio includes: Selecting a low-frequency gain value in the predicted optimized audio as a predicted low-frequency gain value, selecting a number of quantization bits in the predicted optimized audio as a predicted number of quantization bits, and selecting a stereo separation degree in the predicted optimized audio as a predicted stereo separation degree; Selecting a low-frequency gain value in the real optimized audio as a real low-frequency gain value, selecting a number of quantization bits in the real optimized audio as a real number of quantization bits, and selecting a stereo separation degree in the real optimized audio as a real stereo separation degree; The absolute value of the difference between the predicted low-frequency gain value and the actual low-frequency gain value, the absolute value of the difference between the predicted number of quantization bits and the actual number of quantization bits, and the absolute value of the difference between the predicted stereo separation and the actual stereo separation are added to generate a loss value between the predicted optimized audio and the actual optimized audio.

[0010] Exemplarily, the absolute value of the difference between the predicted low-frequency gain value and the actual low-frequency gain value, the absolute value of the difference between the predicted number of quantization bits and the actual number of quantization bits, and the absolute value of the difference between the predicted stereo separation and the actual stereo separation are added to generate a loss value between the predicted optimized audio and the actual optimized audio, including: Selecting the absolute value of the difference between the predicted low-frequency gain value and the actual low-frequency gain value as the first difference value, selecting the absolute value of the difference between the predicted number of quantization bits and the actual number of quantization bits as the second difference value, and selecting the absolute value of the difference between the predicted stereo separation and the actual stereo separation as the third difference value; The absolute value of the difference between the first difference value and the first reference value, the absolute value of the difference between the second difference value and the second reference value, and the absolute value of the difference between the third difference value and the third reference value are added to generate a loss value between the predicted optimized audio and the actual optimized audio.

[0011] Further, adjusting the model parameters of the deep learning model based on the loss value includes: Using a back propagation algorithm, obtaining the gradient of the loss value; Using an optimizer, model parameters of the deep learning model are adjusted according to the gradient.

[0012] Furthermore, when the loss value satisfies a preset condition, the model parameters of the deep learning model are stopped to obtain the adjusted model parameters, and the deep learning model loaded with the adjusted model parameters is selected as the audio optimization model, including: When the loss value is less than a preset value in the preset condition, stop adjusting the model parameters of the deep learning model, obtain the adjusted model parameters, obtain the loading interface of the deep learning model, and obtain the current state of the loading interface; When the current state is an idle state, the adjusted model parameters are loaded into the deep learning model through the loading interface, and the deep learning model loaded with the adjusted model parameters is selected as the audio optimization model.

[0013] Further, the connecting of the Bluetooth headset, obtaining the second music audio to be played, when the music type identifier of the second music audio is a classical music identifier, obtaining the current optimized audio output by the audio optimization model based on the second music audio, and sending the current optimized audio to the Bluetooth headset, comprises: Connecting a Bluetooth headset, obtaining a playlist of a music player, obtaining a second music audio to be played in the playlist, and when the music type identifier of the second music audio is a classical music identifier, performing audio feature extraction on the second music audio to obtain a second feature vector; The second eigenvector is input into the audio optimization model, and the second eigenvector is processed through the audio optimization model to obtain the optimized second music audio, and the optimized second music audio is selected as the current optimized audio, and the current optimized audio is sent to the Bluetooth headset.

[0014] Exemplarily, inputting the second feature vector into the audio optimization model, processing the second feature vector through the audio optimization model to obtain the optimized second music audio, selecting the optimized second music audio as the current optimized audio, and sending the current optimized audio to the Bluetooth headset, includes: Inputting the second feature vector into the audio optimization model, processing the second feature vector through the audio optimization model to obtain the optimized second music audio, and selecting the optimized second music audio as the current optimized audio; Obtaining the original audio signal of the currently optimized audio, obtaining sound effect parameters, and obtaining the intermediate frequency gain value, reverberation time, and compression ratio in the sound effect parameters; Processing the original audio signal according to the intermediate frequency gain value to obtain a first audio signal, processing the first audio signal according to the reverberation time to obtain a second audio signal, and processing the second audio signal according to the compression ratio to obtain a third audio signal; A third audio signal is selected as a target audio signal of the current optimized audio, and the target audio signal is transmitted to the Bluetooth headset.

[0015] Further, in the connecting Bluetooth headset, obtaining the second music audio to be played, when the music type identifier of the second music audio is a classical music identifier, obtaining the current optimized audio output by the audio optimization model based on the second music audio, and sending the current optimized audio to the Bluetooth headset, the optimized audio acquisition method includes: Obtain an upload instruction, execute the upload instruction, and upload the current optimized audio to the music library.

[0016] In a second aspect, a device for obtaining optimized audio based on a Bluetooth headset is provided, comprising: A first acquisition module is used to acquire a first music audio, and when the music type identifier of the first music audio is a classical music identifier, acquire a real optimized audio corresponding to the first music audio; A second acquisition module is used to input the first music audio into a deep learning model to obtain the predicted optimized audio output by the deep learning model based on the first music audio; A third acquisition module, used to obtain a loss value between the predicted optimized audio and the real optimized audio; An adjustment module, configured to adjust a model parameter of the deep learning model based on the loss value; A selection module, used for stopping adjusting the model parameters of the deep learning model when the loss value meets a preset condition, obtaining the adjusted model parameters, and selecting the deep learning model loaded with the adjusted model parameters as the audio optimization model; The fourth acquisition module is used to connect to a Bluetooth headset to obtain a second music audio to be played. When the music type identifier of the second music audio is a classical music identifier, the fourth acquisition module is used to obtain the current optimized audio output by the audio optimization model based on the second music audio, and send the current optimized audio to the Bluetooth headset.

[0017] In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method for optimizing audio acquisition when executing the computer program.

[0018] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned method for optimizing audio acquisition are implemented.

[0019] The present application provides a method, apparatus, computer equipment and storage medium for obtaining optimized audio based on Bluetooth headsets, which obtains a first music audio, and when the music type identifier of the first music audio is a classical music identifier, obtains the real optimized audio corresponding to the first music audio; inputs the first music audio into a deep learning model, and obtains the predicted optimized audio output by the deep learning model based on the first music audio; obtains a loss value between the predicted optimized audio and the real optimized audio; based on the loss value, adjusts the model parameters of the deep learning model; when the loss value meets a preset condition, stops adjusting the model parameters of the deep learning model, obtains the adjusted model parameters, and selects the deep learning model loaded with the adjusted model parameters as the audio optimization model; connects a Bluetooth headset to obtain a second music audio to be played, and when the music type identifier of the second music audio is a classical music identifier, obtains the real optimized audio corresponding to the first music audio; inputs the first music audio into a deep learning model, and obtains the predicted optimized audio output by the deep learning model based on the first music audio; obtains the loss value between the predicted optimized audio and the real optimized audio; based on the loss value, adjusts the model parameters of the deep learning model; when the loss value meets a preset condition, stops adjusting the model parameters of the deep learning model, obtains the adjusted model parameters, and selects the deep learning model loaded with the adjusted model parameters as the audio optimization model; connects the Bluetooth headset to obtain a second music audio to be played, and when the music type identifier of the second music audio is a classical music identifier, obtains the real optimized audio corresponding to the first music audio; inputs the first music audio into a deep learning model, and obtains the predicted optimized audio output by the deep learning model based on the first music audio; obtains the predicted optimized audio and the real optimized audio; obtains the predicted optimized audio and the real optimized audio; adjusts the model parameters of the deep learning model based on the loss value; when the loss value meets a preset condition, stops adjusting the model parameters of the deep learning model, obtains the adjusted model parameters, and selects the deep learning model loaded with the adjusted model parameters as the audio optimization model; When the type identifier is a classical music identifier, the current optimized audio output by the audio optimization model based on the second music audio is obtained, and the current optimized audio is sent to the Bluetooth headset. The beneficial effects are in two aspects. On the one hand, the Bluetooth headset is connected to obtain the second music audio to be played. When the music type identifier of the second music audio is a classical music identifier, the current optimized audio output by the audio optimization model based on the second music audio is obtained, and the current optimized audio is sent to the Bluetooth headset. Since manual acquisition is not required, the acquisition time of the current optimized audio is reduced, which is conducive to improving the acquisition efficiency of the current optimized audio. On the other hand, the target audio signal of the second music audio is obtained, and the target audio signal is sent to the Bluetooth headset. Since it is automatically sent, the sending time of the target audio signal is reduced, which is conducive to improving the sending efficiency of the target audio signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0021] Figure 1 is a schematic diagram of an application environment of an optimized audio acquisition method in an embodiment of the present invention; Figure 2 A schematic flow chart of a method for optimizing audio acquisition provided by an embodiment of the present invention; Figure 3 yes Figure 2 A schematic flow chart of a specific implementation of step S23; Figure 4 yes Figure 2 A schematic flow chart of a specific implementation of step S25; Figure 5 yes Figure 2 A schematic flow chart of a specific implementation of step S26; Figure 6 is a structural schematic diagram of an optimized audio acquisition device in one embodiment of the present invention; Figure 7 It is a structural diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] See also Figure 1 , Figure 1 FIG. 1 is a schematic diagram of an application environment of an optimized audio acquisition method according to an embodiment of the present invention. The optimized audio acquisition method provided by the embodiment of the present invention can be applied in the following embodiments: Figure 1 In the application environment, the computer device is connected to the Bluetooth headset.

[0024] Acquire a first music audio, and when the music type identifier of the first music audio is a classical music identifier, acquire a real optimized audio corresponding to the first music audio; Inputting the first music audio into a deep learning model, and obtaining a predicted optimized audio output by the deep learning model based on the first music audio; Obtaining a loss value between the predicted optimized audio and the real optimized audio; Based on the loss value, adjusting model parameters of the deep learning model; When the loss value meets a preset condition, stop adjusting the model parameters of the deep learning model, obtain the adjusted model parameters, and select the deep learning model loaded with the adjusted model parameters as the audio optimization model; Connect a Bluetooth headset and obtain a second music audio to be played. When the music type identifier of the second music audio is a classical music identifier, obtain the current optimized audio output by the audio optimization model based on the second music audio, and send the current optimized audio to the Bluetooth headset.

[0025] Wherein, the currently optimized audio is sent to the Bluetooth headset, the sender of the currently optimized audio is a computer device, and the receiver of the currently optimized audio is the Bluetooth headset.

[0026] Among them, computer equipment includes but is not limited to servers, desktops, laptops, tablets, and mobile phones.

[0027] Among them, the server can be equipped with a Bluetooth module. The Bluetooth module is a chip basic circuit collection with integrated Bluetooth function, which is used for wireless network communication. After the server is equipped with the Bluetooth module, the server can connect to a Bluetooth headset through the Bluetooth module.

[0028] Among them, the desktop computer can be equipped with a Bluetooth module. The Bluetooth module is a chip basic circuit collection with integrated Bluetooth function, which is used for wireless network communication. After the desktop computer is equipped with the Bluetooth module, the desktop computer can connect to a Bluetooth headset through the Bluetooth module.

[0029] In the scheme implemented by the above-mentioned optimized audio acquisition method, device, equipment and medium, the beneficial effects are in two aspects. On the one hand, a Bluetooth headset is connected to obtain a second music audio to be played. When the music type identifier of the second music audio is a classical music identifier, the current optimized audio output by the audio optimization model based on the second music audio is obtained, and the current optimized audio is sent to the Bluetooth headset. Since manual acquisition is not required, the acquisition time of the current optimized audio is reduced, which is conducive to improving the acquisition efficiency of the current optimized audio. On the other hand, the target audio signal of the second music audio is obtained, and the target audio signal is sent to the Bluetooth headset. Since it is automatically sent, the sending time of the target audio signal is reduced, which is conducive to improving the sending efficiency of the target audio signal.

[0030] The present invention is described in detail below through specific embodiments. Figure 2 , Figure 2 A flowchart of a method for optimizing audio acquisition provided by an embodiment of the present invention includes the following steps: S21, obtaining a first music audio, and when the music type identifier of the first music audio is a classical music identifier, obtaining a real optimized audio corresponding to the first music audio; The first music audio is music audio used for training.

[0031] Among them, the real optimized audio is the first music audio after optimization.

[0032] The music type identifier is an identifier for identifying the music type.

[0033] Among them, music type identification includes classical music identification and non-classical music identification.

[0034] Among them, the classical music logo and the non-classical music logo are different logos.

[0035] Among them, the classical music logo is the logo of classical music.

[0036] Among them, the non-classical music logo is the logo for non-classical music.

[0037] Among them, non-classical music includes pop music and rock music.

[0038] Among them, classical music does not have the interference of the complex element of human voice, which avoids the additional computing burden caused by the complex interaction between human voice and instrument sound, thus reducing the amount of data to be processed, which is conducive to improving the processing speed of the deep learning model. However, pop music and rock music have the interference of the complex element of human voice, which increases the amount of data to be processed, so it is necessary to identify the music type of the first music audio according to the music type identifier of the first music audio.

[0039] Among them, when the music type identifier of the first music audio is a classical music identifier, it means that the audio content of the first music audio is composed or performed according to the style and rules of classical music, and it is necessary to obtain the real optimized audio corresponding to the first music audio.

[0040] Among them, when the music type identifier of the first music audio is not a classical music identifier, it means that the audio content of the first music audio is not created or performed according to the style and rules of classical music, and no processing is required. Therefore, when the music type identifier of the first music audio is not a classical music identifier, the first music audio is not processed.

[0041] Exemplarily, the acquiring of the first music audio, when the music type identifier of the first music audio is a classical music identifier, acquiring the real optimized audio corresponding to the first music audio includes: Acquire a first music audio, and when the music type identifier of the first music audio is a classical music identifier, acquire a plurality of test audios corresponding to the first music audio; Obtain the current storage space, current rating, current playback frequency, and current download count of the test audio; Adding the ratio of the current storage space to the preset storage space, the ratio of the current score to the preset score, the ratio of the current playback frequency to the current playback frequency, and the ratio of the current number of downloads to the preset number of downloads to generate a recommendation coefficient corresponding to the test audio; Obtain the recommendation coefficient corresponding to each of the test audios, and select the test audio with the largest recommendation coefficient as the true optimized audio corresponding to the first music audio.

[0042] Exemplarily, the ratio of the current storage space to the preset storage space, the ratio of the current score to the preset score, the ratio of the current playback frequency to the current playback frequency, and the ratio of the current download count to the preset download count are added to generate a recommendation coefficient corresponding to the test audio, including: Selecting a ratio of the current storage space to the preset storage space as a first ratio, selecting a ratio of the current score to the preset score as a second ratio, selecting a ratio of the current playback frequency to the current playback frequency as a third ratio, and selecting a ratio of the current number of downloads to the preset number of downloads as a fourth ratio; According to the recommendation coefficient generation model, adding the minimum value of the first ratio and the second ratio and the minimum value of the third ratio and the fourth ratio to generate a recommendation coefficient corresponding to the test audio; The recommended coefficient generation model is: S=min(S1, S2) + min(S3, S4); Wherein, S is the recommended coefficient, S1 is the first ratio, S2 is the second ratio, S3 is the third ratio, and S4 is the fourth ratio; Among them, the recommendation coefficient generation model is a generation model of the recommendation coefficient.

[0043] Among them, the min function is used to find the minimum value from S1 and S2, and to find the minimum value from S3 and S4. In practical applications, the minimum value can reveal the factors that limit the recommendation coefficient. Therefore, adaptive adjustment is required to increase the minimum value of S1 and S2, and increase the minimum value of S3 and S4, thereby improving the recommendation coefficient.

[0044] For example, between the first ratio and the second ratio, if the first ratio is the minimum value, it means that the current storage space is insufficient and the current storage space needs to be reduced; if the second ratio is the minimum value, it means that the current score is insufficient and the current score needs to be increased.

[0045] For example, among the third ratio and the fourth ratio, if the third ratio is the minimum value, it means that the current playback frequency is insufficient and the current playback frequency needs to be increased; if the fourth ratio is the minimum value, it means that the current download times are insufficient and the current download times need to be increased.

[0046] Among them, the recommendation coefficient combines the first ratio, the second ratio, the third ratio, and the fourth ratio, so that the recommendation degree of the test audio can be evaluated as a whole.

[0047] The first music audio and the real optimized audio are used as a set of training samples. When constructing a large number of training samples, the real optimized audio is not randomly selected, but is obtained based on the current storage space, current rating, current playback frequency, and current download count. This ensures that the real optimized audio is widely recognized in terms of its effect.

[0048] For the sake of illustration, an example is given below: For example, after the first music audio is optimized by the engineer, test audio 1, test audio 2, test audio 3, test audio 4, and test audio 5 are obtained; Through the recommendation coefficient mechanism, obtain the recommendation coefficients of test audio 1, test audio 2, test audio 3, test audio 4, and test audio 5; Among test audio 1, test audio 2, test audio 3, test audio 4, and test audio 5, the test audio with the highest recommendation coefficient is selected as the real optimized audio.

[0049] When the recommendation coefficient of music audio 1 is the highest, music audio 1 is selected as the real optimized audio.

[0050] When the recommendation coefficient of music audio 2 is the highest, music audio 2 is selected as the real optimized audio.

[0051] When the recommendation coefficient of music audio 3 is the highest, music audio 3 is selected as the real optimized audio.

[0052] When the recommendation coefficient of music audio 5 is the highest, music audio 5 is selected as the real optimized audio.

[0053] Among them, this method can screen out real optimized audio that has excellent technical performance and is deeply loved by users.

[0054] S22, inputting the first music audio into a deep learning model, and obtaining a predicted optimized audio output by the deep learning model based on the first music audio; Among them, the real optimized audio provides a clear training target for the deep learning model. During the training process, the deep learning model will try to extract features from the first music audio and learn how to map these features to the real optimized audio. The first music audio contains various music features. The deep learning model learns the features learned by the first music audio and outputs the predicted optimized audio. The deep learning model can gradually master how to extract useful information from complex audio data, thereby generating predicted optimized audio similar to the real optimized audio.

[0055] The step of inputting the first music audio into a deep learning model and obtaining a predicted optimized audio output by the deep learning model based on the first music audio includes: Performing audio feature extraction on the first music audio to obtain a first feature vector; The first feature vector is input into a deep learning model, and the first feature vector is processed by the deep learning model to obtain predicted optimized audio.

[0056] Among them, the deep learning model can adopt a pre-trained model.

[0057] Among them, the use of pre-trained models can significantly reduce training time, because pre-trained models learn basic and generally applicable knowledge on a wide range of data sets, and only fine-tuning needs to be done based on the pre-trained models. The use of pre-trained models in deep learning models can reduce the cost and time of training deep learning models, greatly improving the training efficiency of deep learning models.

[0058] For the sake of illustration, an example is given below: For example, the deep learning model uses the GPT model or the Wavenet model.

[0059] The Chinese name of the Wavenet model is: wavy network, and the English name of the Wavenet model is: WaveNet. The Wavenet model significantly improves the ability of the Wavenet model to capture the timing dependency and frequency characteristics of audio data by introducing dilated convolution.

[0060] The Chinese name of the GPT model is: Generative Pre-trained Transformer model, and the English name of the GPT model is: Generative Pre-trained Transformer. Although the GPT model was originally designed for natural language processing tasks, due to the GPT model's powerful generation capabilities and good processing of sequence data, the application scope of the GPT model has been expanded to the field of music generation.

[0061] S23, obtaining a loss value between the predicted optimized audio and the real optimized audio; Among them, the loss value is used to measure the degree of difference between the predicted optimized audio and the real optimized audio; the larger the loss value, the greater the degree of difference between the predicted optimized audio and the real optimized audio, and the smaller the loss value, the smaller the degree of difference between the predicted optimized audio and the real optimized audio.

[0062] S24, adjusting the model parameters of the deep learning model based on the loss value; The model parameters of the deep learning model are adjusted with the goal of minimizing the loss value. The smaller the loss value, the smaller the error between the predicted optimized audio and the real optimized audio, and the stronger the performance of the deep learning model. Therefore, by continuously adjusting the parameters of the deep learning model, the loss value is gradually reduced, thereby improving the performance of the deep learning model.

[0063] Wherein, adjusting the model parameters of the deep learning model based on the loss value includes: Using a back propagation algorithm, obtaining the gradient of the loss value; Using an optimizer, model parameters of the deep learning model are adjusted according to the gradient.

[0064] Among them, the back propagation algorithm is a supervised learning algorithm that uses the chain derivative rule to calculate the gradient of the loss function for each layer of parameters.

[0065] The gradient indicates the direction and size of the model parameter adjustment. Through the gradient, the optimizer can determine which model parameters need to be increased, which model parameters need to be decreased, and the magnitude of the adjustment. The optimizer only uses the gradient of one sample at a time to update the parameters, which helps speed up training.

[0066] S25, when the loss value meets the preset conditions, stop adjusting the model parameters of the deep learning model, obtain the adjusted model parameters, and select the deep learning model loaded with the adjusted model parameters as the audio optimization model; When the loss value meets the preset conditions, stop adjusting the model parameters of the deep learning model to obtain the adjusted model parameters, which can avoid unnecessary waste of resources, reduce training time and improve training efficiency.

[0067] S26, connect the Bluetooth headset, obtain the second music audio to be played, when the music type identifier of the second music audio is a classical music identifier, obtain the current optimized audio output by the audio optimization model based on the second music audio, and send the current optimized audio to the Bluetooth headset.

[0068] Among them, the second music audio is the music audio to be optimized.

[0069] When the music type identifier of the second music audio is not a classical music identifier, the second music audio is not processed.

[0070] The currently optimized audio is the optimized second music audio. After optimization, the sound quality of the second music audio can be significantly improved, so that the user can hear the details in the second music audio more clearly.

[0071] When the music type identifier of the second music audio is a classical music identifier, it means that the audio content of the second music audio is composed or performed according to the style and rules of classical music.

[0072] Among them, the music type identification of the first music audio is a classical music identification, and the music type identification of the second music audio is a classical music identification, which means that the second music audio and the first music audio are different music audios in classical music.

[0073] Among them, the audio optimization model can output the current optimized audio according to the second music audio. This direct and efficient optimization method not only reduces the need for manual intervention, but also ensures the consistency and repeatability of the optimization process, so that the second music audio can be optimized in a short time, thereby greatly improving the optimization efficiency of the second music audio. This method can be applied to the optimization scenario of classical music works. For example, in classical music works, a music audio is randomly selected as the second music audio, and then the second music audio is input into the audio optimization model to obtain the current optimized audio output by the audio optimization model based on the second music audio. In this way, the classical music work can be optimized in a short time, thereby simplifying the optimization process and improving the optimization efficiency of the classical music work.

[0074] Wherein, in the connecting of the Bluetooth headset, obtaining the second music audio to be played, when the music type identifier of the second music audio is a classical music identifier, obtaining the current optimized audio output by the audio optimization model based on the second music audio, and sending the current optimized audio to the Bluetooth headset, the optimized audio acquisition method includes: Obtain an upload instruction, execute the upload instruction, and upload the current optimized audio to the music library.

[0075] Among them, the music library is a database for storing music audio.

[0076] Among them, uploading the current optimized audio to the music library can realize the reuse of the current optimized audio, which not only greatly reduces the workload of repeatedly producing the current optimized audio, but also significantly improves the utilization efficiency of the current optimized audio.

[0077] This application is applied to a laptop computer. The process of connecting a laptop computer to a Bluetooth headset is described as follows: 1. The Bluetooth headset enters pairing mode: The Bluetooth headset needs to be set to a detectable state, that is, pairing mode. This is usually achieved by long pressing a button on the Bluetooth headset until the Bluetooth headset indicator starts flashing, indicating that it has entered pairing mode.

[0078] 2. Turn on Bluetooth on the laptop and search for devices: Next, the user needs to turn on the Bluetooth function on the laptop and start the device search. The laptop will start scanning for available Bluetooth devices around it.

[0079] 3. Laptop sends connection request: When the laptop detects a Bluetooth headset in pairing mode, it will display the name or logo of the Bluetooth headset on the screen. At this point, the user can select the Bluetooth headset and click Connect. This operation actually sends a connection request from the laptop to the Bluetooth headset.

[0080] 4. The Bluetooth headset responds to the connection request: After receiving the connection request from the laptop, the Bluetooth headset will verify and confirm the connection. For some Bluetooth headsets, the user may be required to confirm the pairing code or perform other authentication steps by pressing a button on the Bluetooth headset or prompting on the laptop.

[0081] 5. Connection establishment: Once the Bluetooth headset confirms the connection request, the Bluetooth connection between the Bluetooth headset and the laptop is established. At this time, the laptop obtains the second music audio to be played, and when the music type identifier of the second music audio is a classical music identifier, the current optimized audio output by the audio optimization model based on the second music audio is obtained, and the current optimized audio is sent to the Bluetooth headset, and the user can start listening to music through the Bluetooth headset.

[0082] In the embodiment of the present invention, the beneficial effects are in two aspects. On the one hand, a Bluetooth headset is connected to obtain a second music audio to be played. When the music type identifier of the second music audio is a classical music identifier, the current optimized audio output by the audio optimization model based on the second music audio is obtained, and the current optimized audio is sent to the Bluetooth headset. Since manual acquisition is not required, the acquisition time of the current optimized audio is reduced, which is beneficial to improving the acquisition efficiency of the current optimized audio. On the other hand, the target audio signal of the second music audio is obtained, and the target audio signal is sent to the Bluetooth headset. Since it is automatically sent, the sending time of the target audio signal is reduced, which is beneficial to improving the sending efficiency of the target audio signal.

[0083] See also Figure 3 , Figure 3 yes Figure 2 A specific implementation flow diagram of step S23 is described in detail as follows: S31, selecting a low-frequency gain value in the predicted optimized audio as a predicted low-frequency gain value, selecting a number of quantization bits in the predicted optimized audio as a predicted number of quantization bits, and selecting a stereo separation degree in the predicted optimized audio as a predicted stereo separation degree; S32, selecting a low-frequency gain value in the real optimized audio as a real low-frequency gain value, selecting a number of quantization bits in the real optimized audio as a real number of quantization bits, and selecting a stereo separation degree in the real optimized audio as a real stereo separation degree; S33, adding the absolute value of the difference between the predicted low-frequency gain value and the actual low-frequency gain value, the absolute value of the difference between the predicted number of quantization bits and the actual number of quantization bits, and the absolute value of the difference between the predicted stereo separation and the actual stereo separation, to generate a loss value between the predicted optimized audio and the actual optimized audio.

[0084] The low-frequency gain value refers to the amount of gain added when the audio signal is amplified or enhanced in the low-frequency band, usually expressed in decibels. The low-frequency band refers to the low-frequency range of 20Hz to 200Hz.

[0085] Among them, the low-frequency gain value can enhance the bass in the target audio signal, making the bass richer and fuller, thereby improving the overall sound quality and the realism of the sound effect. In the target audio signal, the bass is mainly composed of the sound of large instruments such as cello, double bass and organ. The low-frequency gain value can strengthen the sound of cello, double bass and organ, making the sound of these instruments fuller, deeper and more expressive, thereby enhancing the overall layering and emotional depth of classical music.

[0086] Among them, when the absolute value of the difference between the predicted low-frequency gain value and the actual low-frequency gain value is smaller, that is, the difference between the low-frequency gain value of the predicted optimized audio and the low-frequency gain value of the actual optimized audio is smaller, it means that the predicted optimized audio and the actual optimized audio are closer in enhancing the sounds produced by cellos, double basses and organs.

[0087] The number of quantization bits determines the dynamic range of the audio signal after digitization, that is, the ratio between the loudest sound and the quietest sound. A higher number of quantization bits can provide a larger dynamic range, making the audio signal smoother when the volume changes.

[0088] Therefore, the smaller the difference in quantization bits, the closer the predicted optimized audio and the real optimized audio are in recording the range of sound from the weakest to the strongest, as well as the rich details and layering of sound within this range. The difference in auditory effect between the predicted optimized audio and the real optimized audio is also smaller.

[0089] Among them, stereo separation is an important indicator to measure the difference between the left and right channels, reflecting the spatial performance of the audio signal.

[0090] The step of selecting the stereo separation degree in the predicted optimized audio as the predicted stereo separation degree comprises: The predicted optimized audio is imported into the audio testing software, and the predicted optimized audio is processed by the audio testing software to generate the stereo separation in the predicted optimized audio.

[0091] The step of selecting the stereo separation degree in the real optimized audio as the real stereo separation degree comprises: Import the real optimized audio into the audio testing software, process the real optimized audio through the audio testing software, and generate the stereo separation in the real optimized audio.

[0092] Among them, the smaller the absolute value of the difference between the predicted stereo separation and the real stereo separation, that is, the smaller the difference between the stereo separation of the predicted optimized audio and the stereo separation of the real optimized audio, the less significant the difference in effect between the predicted optimized audio and the real optimized audio in creating a stereo environment, presenting sound positioning and spatial layering, and thus the smaller the auditory difference felt by the user when listening.

[0093] Among them, since the loss value comprehensively considers the absolute value of the difference between the predicted low-frequency gain value and the actual low-frequency gain value, the absolute value of the difference between the predicted number of quantization bits and the actual number of quantization bits, and the absolute value of the difference between the predicted stereo separation and the actual stereo separation, the loss value can provide comprehensive and detailed feedback to the deep learning model.

[0094] Exemplarily, the absolute value of the difference between the predicted low-frequency gain value and the actual low-frequency gain value, the absolute value of the difference between the predicted number of quantization bits and the actual number of quantization bits, and the absolute value of the difference between the predicted stereo separation and the actual stereo separation are added to generate a loss value between the predicted optimized audio and the actual optimized audio, including: Selecting the absolute value of the difference between the predicted low-frequency gain value and the actual low-frequency gain value as the first difference value, selecting the absolute value of the difference between the predicted number of quantization bits and the actual number of quantization bits as the second difference value, and selecting the absolute value of the difference between the predicted stereo separation and the actual stereo separation as the third difference value; The absolute value of the difference between the first difference value and the first reference value, the absolute value of the difference between the second difference value and the second reference value, and the absolute value of the difference between the third difference value and the third reference value are added to generate a loss value between the predicted optimized audio and the actual optimized audio.

[0095] When the gaps between the first difference value, the second difference value, and the third difference value are large, the relative sizes and degree of differences of the first difference value, the second difference value, and the third difference value under different benchmarks can be more accurately reflected by introducing their respective first reference values, second reference values, and third reference values. This can not only eliminate the magnitude differences between the first difference value, the second difference value, and the third difference value, but also make the comparison more fair, thereby increasing the accuracy of the analysis.

[0096] Among them, the loss value, as a comprehensive indicator, can simultaneously reflect the changes in the first difference value, the second difference value, and the third difference value relative to their respective first reference value, the second reference value, and the third reference value, so that data of different magnitudes or different backgrounds can be compared on the same scale, which ensures the fairness and objectivity of the analysis results and is conducive to improving the efficiency and accuracy of the analysis.

[0097] Exemplarily, the absolute value of the difference between the first difference value and the first reference value, the absolute value of the difference between the second difference value and the second reference value, and the absolute value of the difference between the third difference value and the third reference value are added to generate a loss value between the predicted optimized audio and the real optimized audio, including: According to the loss value generation model, the absolute value of the difference between the first difference value and the first reference value, the absolute value of the difference between the second difference value and the second reference value, and the absolute value of the difference between the third difference value and the third reference value are added to generate a loss value between the predicted optimized audio and the real optimized audio; The loss value generation model is: L=max(L1, L2)+ max(L2, L3); Among them, L is the loss value, L1 is the absolute value of the difference between the first difference value and the first reference value, L2 is the absolute value of the difference between the second difference value and the second reference value, and L3 is the absolute value of the difference between the third difference value and the third reference value.

[0098] Among them, the max function is used to find the maximum value from L1 and L2, L2 and L3, and can determine the factors that affect the loss value in L1 and L2, L2 and L3. Therefore, adaptive adjustment is required to reduce the maximum value in L1 and L2, and reduce the maximum value in L2 and L3, thereby reducing the loss value.

[0099] For ease of explanation, the details are as follows: Among them, among L1 and L2, if L1 is the maximum value, it means that the difference between the predicted low-frequency gain value and the actual low-frequency gain value needs to be reduced; if L2 is the maximum value, it means that the difference between the predicted low-frequency gain value and the actual low-frequency gain value needs to be reduced.

[0100] Among them, among L2 and L3, if L2 is the maximum value, it means that the difference between the predicted low-frequency gain value and the actual low-frequency gain value needs to be reduced; if L3 is the maximum value, it means that the difference between the predicted low-frequency gain value and the actual low-frequency gain value needs to be reduced.

[0101] Among them, the loss value integrates the absolute value of the difference between the first difference value and the first reference value, the absolute value of the difference between the second difference value and the second reference value, and the absolute value of the difference between the third difference value and the third reference value, so that the degree of difference between the predicted optimized audio and the real optimized audio can be evaluated as a whole.

[0102] In an embodiment of the present invention, compared with a single loss value, the loss values ​​obtained from multiple aspects can capture the deficiencies of the loss value in different aspects, thereby guiding the deep learning model to perform more accurate parameter adjustments during the training process. This refined optimization is not only conducive to improving the overall performance of the deep learning model, but also can enhance the adaptability of the deep learning model in complex scenarios.

[0103] See also Figure 4 , Figure 4 yes Figure 2 A specific implementation flow diagram of step S25 is described in detail as follows: S41, when the loss value is less than a preset value in the preset condition, stop adjusting the model parameters of the deep learning model, obtain the adjusted model parameters, obtain the loading interface of the deep learning model, and obtain the current state of the loading interface; S42, when the current state is an idle state, the adjusted model parameters are loaded into the deep learning model through the loading interface, and the deep learning model loaded with the adjusted model parameters is selected as the audio optimization model.

[0104] Among them, when the loss value is less than the preset value in the preset condition, it means that the error between the predicted optimized audio and the real optimized audio of the deep learning model loaded with the adjusted model parameters has reached a relatively stable low level, which reflects that the deep learning model using the trained model parameters has fully learned the inherent laws and characteristics of the real optimized audio. Therefore, selecting the deep learning model loaded with the adjusted model parameters as the audio optimization model can ensure that the audio optimization model has good generalization ability.

[0105] In an embodiment of the present invention, the deep learning model loaded with the adjusted model parameters is selected as the audio optimization model. The adjusted model parameters are conducive to the audio optimization model to better capture the potential rules and characteristics in the data, so that the audio optimization model can make more accurate and reliable predictions when processing the second music audio.

[0106] See also Figure 5 , Figure 5 yes Figure 2 A specific implementation flow diagram of step S26 is described in detail as follows: S51, connecting a Bluetooth headset, obtaining a playlist of a music player, obtaining a second music audio to be played in the playlist, and when the music type identifier of the second music audio is a classical music identifier, performing audio feature extraction on the second music audio to obtain a second feature vector; When the music player starts playing a playlist, the second music audio to be played is determined according to the order in the playlist.

[0107] S52, input the second eigenvector into the audio optimization model, process the second eigenvector through the audio optimization model to obtain the optimized second music audio, select the optimized second music audio as the current optimized audio, and send the current optimized audio to the Bluetooth headset.

[0108] Exemplarily, inputting the second feature vector into the audio optimization model, processing the second feature vector through the audio optimization model to obtain the optimized second music audio, selecting the optimized second music audio as the current optimized audio, and sending the current optimized audio to the Bluetooth headset, includes: Inputting the second feature vector into the audio optimization model, processing the second feature vector through the audio optimization model to obtain the optimized second music audio, and selecting the optimized second music audio as the current optimized audio; Obtaining the original audio signal of the currently optimized audio, obtaining sound effect parameters, and obtaining the intermediate frequency gain value, reverberation time, and compression ratio in the sound effect parameters; Processing the original audio signal according to the intermediate frequency gain value to obtain a first audio signal, processing the first audio signal according to the reverberation time to obtain a second audio signal, and processing the second audio signal according to the compression ratio to obtain a third audio signal; A third audio signal is selected as a target audio signal of the current optimized audio, and the target audio signal is transmitted to the Bluetooth headset.

[0109] Exemplarily, the method includes processing the original audio signal according to the intermediate frequency gain value to obtain a first audio signal, processing the first audio signal according to the reverberation time to obtain a second audio signal, and processing the second audio signal according to the compression ratio to obtain a third audio signal, including: The original audio signal is input into an equalizer, and a first audio signal output by the equalizer based on the intermediate frequency gain value is obtained; the first audio signal is input into a reverberator, and a second audio signal output by the reverberator based on the reverberation time is obtained; the second audio signal is input into a compressor, and a third audio signal output by the compressor based on the compression ratio is obtained.

[0110] The first audio signal is optimized for the intermediate frequency gain value, and the second audio signal is optimized for the intermediate frequency gain value and the reverberation time.

[0111] The third audio signal, i.e., the target audio signal, has been optimized for the intermediate frequency gain value, the reverberation time, and the compression ratio. Therefore, the target audio signal is conducive to satisfying the user's personalized needs for sound effects and protecting the user's hearing.

[0112] For ease of explanation, the details are as follows; The original audio signal of the current optimized audio can be further optimized through the intermediate frequency gain value and the reverberation time to obtain the target audio signal of the current optimized audio.

[0113] The mid-frequency gain value is the gain in the mid-frequency range, usually measured in decibels (dB). The mid-frequency range refers to 200Hz to 2000Hz.

[0114] Among them, the reverberation time can significantly enhance the spatial and three-dimensional sense of the target audio signal, creating a more realistic and rich auditory environment, such as a concert hall, theater or outdoor space, making the sound in the target audio signal more in line with the actual scene, enhancing the user's sense of immersion and auditory experience.

[0115] Therefore, the target audio signal is optimized in terms of intermediate frequency gain value and reverberation time, which is conducive to meeting the user's personalized needs for sound effects.

[0116] Among them, the compression ratio is an important parameter in audio processing. The compression ratio determines how the compressor will reduce the volume of the excess part when the volume of the audio signal exceeds the set threshold.

[0117] The compression ratio is usually expressed as input increase: output increase, such as 2:1, 4:1.

[0118] Preferably, the compression ratio is 2:1.

[0119] Among them, the 2:1 ratio: when the volume of the original audio signal exceeds the set threshold, the compressor will halve the volume of the excess part, that is, for every 1dB increase in input volume, the output volume will only increase by 0.5dB. This ratio is suitable for scenes that need to maintain a certain volume stability, such as solo instrument performances, to avoid sudden volume increases that may cause discomfort to users, which is beneficial to protecting users' hearing.

[0120] Among them, the 4:1 ratio means that when the volume of the original audio signal exceeds the set threshold, the output volume only increases by 0.25dB for every 1dB increase in input volume.

[0121] Therefore, the compression ratio of the target audio signal is optimized, which can avoid discomfort to the user due to a sudden increase in volume and is beneficial to protecting the user's hearing.

[0122] Therefore, the target audio signal is beneficial to both meeting the user's personalized needs for sound effects and protecting the user's hearing.

[0123] In an embodiment of the present invention, the second eigenvector is input into the audio optimization model, and the second eigenvector is processed by the audio optimization model to obtain an optimized second music audio, and the optimized second music audio is selected as the current optimized audio. Since manual acquisition is not required, the acquisition time of the current optimized audio is reduced, which is beneficial to improving the acquisition efficiency of the current optimized audio.

[0124] See also Figure 6 , Figure 6 is a schematic diagram of a structure of an optimized audio acquisition device in one embodiment of the present invention. Figure 6 As shown, the optimized audio acquisition device includes a first acquisition module 101, a second acquisition module 102, a third acquisition module 103, an adjustment module 104, a selection module 105, and a fourth acquisition module 106. The detailed description of each functional module is as follows: The first acquisition module 101 is used to acquire a first music audio, and when the music type identifier of the first music audio is a classical music identifier, acquire a real optimized audio corresponding to the first music audio; A second acquisition module 102 is used to input the first music audio into a deep learning model to obtain a predicted optimized audio output by the deep learning model based on the first music audio; A third acquisition module 103 is used to obtain a loss value between the predicted optimized audio and the real optimized audio; An adjustment module 104, configured to adjust a model parameter of the deep learning model based on the loss value; A selection module 105 is used to stop adjusting the model parameters of the deep learning model when the loss value meets a preset condition, obtain the adjusted model parameters, and select the deep learning model loaded with the adjusted model parameters as the audio optimization model; The fourth acquisition module 106 is used to connect to a Bluetooth headset to obtain a second music audio to be played. When the music type identifier of the second music audio is a classical music identifier, the current optimized audio output by the audio optimization model based on the second music audio is obtained, and the current optimized audio is sent to the Bluetooth headset.

[0125] In the embodiment of the present invention, the beneficial effects are in two aspects. On the one hand, a Bluetooth headset is connected to obtain a second music audio to be played. When the music type identifier of the second music audio is a classical music identifier, the current optimized audio output by the audio optimization model based on the second music audio is obtained, and the current optimized audio is sent to the Bluetooth headset. Since manual acquisition is not required, the acquisition time of the current optimized audio is reduced, which is beneficial to improving the acquisition efficiency of the current optimized audio. On the other hand, the target audio signal of the second music audio is obtained, and the target audio signal is sent to the Bluetooth headset. Since it is automatically sent, the sending time of the target audio signal is reduced, which is beneficial to improving the sending efficiency of the target audio signal.

[0126] For the specific limitations on the optimized audio acquisition device, please refer to the limitations on the optimized audio acquisition method above, which will not be repeated here.

[0127] Each module in the above-mentioned optimized audio acquisition device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0128] See also Figure 7 , Figure 7 1 is another structural diagram of a computer device in one embodiment of the present invention. In one embodiment, a computer device is provided. The computer device is a server device or a client device. The internal structure diagram thereof can be as shown in FIG. Figure 7As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external network. When the computer program is executed by the processor, the functions or steps of an optimized audio acquisition method based on a Bluetooth headset can be implemented.

[0129] In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor.

[0130] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant description of the aforementioned method embodiment. In order to avoid repetition, they will not be described one by one here.

[0131] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a graphics processing unit (GPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0132] The above description and the accompanying drawings fully illustrate the embodiments of the present disclosure so that those skilled in the art can practice them. Other embodiments may include structural, logical, electrical, process and other changes. The embodiments represent only possible variations. Unless explicitly required, individual components and functions are optional, and the order of operations may vary. Parts and sub-samples of some embodiments may be included in or replace parts and sub-samples of other embodiments. Moreover, the terms used in this application are only used to describe the embodiments and are not used to limit the claims. In this article, each embodiment may focus on the differences from other embodiments, and the same and similar parts between the various embodiments may refer to each other. For the methods, products, etc. disclosed in the embodiments, if they correspond to the method part disclosed in the embodiments, then the relevant parts can refer to the description of the method part.

[0133] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software may depend on the specific application and design constraints of the technical solution. Technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the embodiments of the present disclosure. Technicians can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0134] In the embodiments disclosed herein, the disclosed methods and products (including but not limited to devices, equipment, etc.) can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units can be only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some sub-samples can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between each other shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to implement this embodiment. In addition, each functional unit in the embodiment of the present disclosure may be integrated in a processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit.

[0135] The flowchart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to the embodiment of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. In the description corresponding to the flowchart and the block diagram in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in a different order from the order disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.

Claims

1. A method for obtaining optimized audio based on a Bluetooth headset, characterized in that: include: Acquire a first music audio, and when the music type identifier of the first music audio is a classical music identifier, acquire a real optimized audio corresponding to the first music audio; Inputting the first music audio into a deep learning model, and obtaining a predicted optimized audio output by the deep learning model based on the first music audio; Obtaining a loss value between the predicted optimized audio and the real optimized audio; Based on the loss value, adjusting model parameters of the deep learning model; When the loss value meets a preset condition, stop adjusting the model parameters of the deep learning model, obtain the adjusted model parameters, and select the deep learning model loaded with the adjusted model parameters as the audio optimization model; Connect a Bluetooth headset and obtain a second music audio to be played. When the music type identifier of the second music audio is a classical music identifier, obtain the current optimized audio output by the audio optimization model based on the second music audio, and send the current optimized audio to the Bluetooth headset.

2. The optimized audio acquisition method according to claim 1, characterized in that: The step of inputting the first music audio into a deep learning model and obtaining a predicted optimized audio output by the deep learning model based on the first music audio includes: Performing audio feature extraction on the first music audio to obtain a first feature vector; The first feature vector is input into a deep learning model, and the first feature vector is processed by the deep learning model to obtain predicted optimized audio.

3. The optimized audio acquisition method according to claim 1, characterized in that: The obtaining of the loss value between the predicted optimized audio and the real optimized audio includes: Selecting a low-frequency gain value in the predicted optimized audio as a predicted low-frequency gain value, selecting a number of quantization bits in the predicted optimized audio as a predicted number of quantization bits, and selecting a stereo separation degree in the predicted optimized audio as a predicted stereo separation degree; Selecting a low-frequency gain value in the real optimized audio as a real low-frequency gain value, selecting a number of quantization bits in the real optimized audio as a real number of quantization bits, and selecting a stereo separation degree in the real optimized audio as a real stereo separation degree; The absolute value of the difference between the predicted low-frequency gain value and the actual low-frequency gain value, the absolute value of the difference between the predicted number of quantization bits and the actual number of quantization bits, and the absolute value of the difference between the predicted stereo separation and the actual stereo separation are added to generate a loss value between the predicted optimized audio and the actual optimized audio.

4. The optimized audio acquisition method according to claim 1, characterized in that: The adjusting the model parameters of the deep learning model based on the loss value includes: Using a back propagation algorithm, obtaining the gradient of the loss value; Using an optimizer, model parameters of the deep learning model are adjusted according to the gradient.

5. The optimized audio acquisition method according to claim 1, characterized in that: When the loss value satisfies a preset condition, the model parameters of the deep learning model are stopped to obtain the adjusted model parameters, and the deep learning model loaded with the adjusted model parameters is selected as the audio optimization model, including: When the loss value is less than a preset value in the preset condition, stop adjusting the model parameters of the deep learning model, obtain the adjusted model parameters, obtain the loading interface of the deep learning model, and obtain the current state of the loading interface; When the current state is an idle state, the adjusted model parameters are loaded into the deep learning model through the loading interface, and the deep learning model loaded with the adjusted model parameters is selected as the audio optimization model.

6. The optimized audio acquisition method according to claim 1, characterized in that: The method of connecting a Bluetooth headset, obtaining a second music audio to be played, and when the music type identifier of the second music audio is a classical music identifier, obtaining a current optimized audio output by the audio optimization model based on the second music audio, and sending the current optimized audio to the Bluetooth headset includes: Connecting a Bluetooth headset, obtaining a playlist of a music player, obtaining a second music audio to be played in the playlist, and when the music type identifier of the second music audio is a classical music identifier, performing audio feature extraction on the second music audio to obtain a second feature vector; The second eigenvector is input into the audio optimization model, and the second eigenvector is processed through the audio optimization model to obtain the optimized second music audio, and the optimized second music audio is selected as the current optimized audio, and the current optimized audio is sent to the Bluetooth headset.

7. The optimized audio acquisition method according to claim 1, characterized in that: In the connecting of the Bluetooth headset, obtaining the second music audio to be played, when the music type identifier of the second music audio is a classical music identifier, obtaining the current optimized audio output by the audio optimization model based on the second music audio, and sending the current optimized audio to the Bluetooth headset, the optimized audio acquisition method includes: Obtain an upload instruction, execute the upload instruction, and upload the current optimized audio to the music library.

8. An optimized audio acquisition device based on Bluetooth headset, characterized in that: include: A first acquisition module is used to acquire a first music audio, and when the music type identifier of the first music audio is a classical music identifier, acquire a real optimized audio corresponding to the first music audio; A second acquisition module is used to input the first music audio into a deep learning model to obtain the predicted optimized audio output by the deep learning model based on the first music audio; A third acquisition module, used to obtain a loss value between the predicted optimized audio and the real optimized audio; An adjustment module, configured to adjust a model parameter of the deep learning model based on the loss value; A selection module, used for stopping adjusting the model parameters of the deep learning model when the loss value meets a preset condition, obtaining the adjusted model parameters, and selecting the deep learning model loaded with the adjusted model parameters as the audio optimization model; The fourth acquisition module is used to connect to a Bluetooth headset to obtain a second music audio to be played. When the music type identifier of the second music audio is a classical music identifier, the fourth acquisition module is used to obtain the current optimized audio output by the audio optimization model based on the second music audio, and send the current optimized audio to the Bluetooth headset.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for optimizing audio acquisition according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for optimizing audio acquisition according to any one of claims 1 to 7 are implemented.

Images