Audio playing method and device, equipment and storage medium

Abstract

The invention relates to the technical field of audio playing, and discloses an audio playing method and device, equipment and a storage medium. The method comprises the following steps: receiving a digital audio signal through a wireless receiving module in the earphone, and analyzing the digital audio signal through a generalized Gaussian mixture model to obtain audio feature data; performing audio decoding and sound effect enhancement on the audio feature data to obtain initial audio data; performing audio reconstruction of residual error conversion on the initial audio data to obtain audio reconstruction data; performing digital-analog conversion on the audio reconstruction data to obtain an initial analog audio signal; performing feature alignment of sorting transformation on the initial analog audio signal to obtain a target analog audio signal; and performing signal amplification on the target analog audio signal to obtain an amplified audio signal, performing sound wave conversion on the amplified audio signal through the earphone, and outputting a target audio, thereby improving the audio playing effect of the Bluetooth earphone.

Description

Technical Field

[0001] The present application relates to the field of audio playback technology, and in particular to an audio playback method, device, equipment and storage medium. Background Art

[0002] With the rapid development of mobile devices and wireless technology, users' requirements for sound quality continue to increase, and the demand for lossless audio playback experience is increasing.

[0003] However, due to signal attenuation and interference during wireless transmission and the limitations of the compression algorithm itself, the quality of the audio signal is often impaired, especially when transmitting high-quality audio. These factors work together to make the final audio output often unable to meet the requirements of high-fidelity audio, thus affecting the user's auditory experience. In addition, when processing complex audio signals, existing audio playback technologies often lack effective optimization algorithms to deal with noise and distortion in audio signals, especially in the decoding and reconstruction stages of the signal. Summary of the invention

[0004] The present application provides an audio playback method, apparatus, device and storage medium for improving the audio playback effect of a Bluetooth headset.

[0005] In a first aspect, the present application provides an audio playback method, the audio playback method comprising: The digital audio signal is received through a wireless receiving module in the headset, and the digital audio signal is analyzed through a generalized Gaussian mixture model to obtain audio feature data; Performing audio decoding and sound effect enhancement on the audio feature data to obtain initial audio data; Performing audio reconstruction by residual transformation on the initial audio data to obtain audio reconstruction data; Performing digital-to-analog conversion on the audio reconstruction data to obtain an initial analog audio signal; Performing feature alignment of sorting transformation on the initial analog audio signal to obtain a target analog audio signal; The target analog audio signal is amplified to obtain an amplified audio signal, and the amplified audio signal is converted into a sound wave through the earphone to output the target audio.

[0006] In a second aspect, the present application provides an audio playback device, the audio playback device comprising: A receiving module, used to receive a digital audio signal through a wireless receiving module in the headset, and analyze the digital audio signal through a generalized Gaussian mixture model to obtain audio feature data; A decoding module, used to perform audio decoding and sound effect enhancement on the audio feature data to obtain initial audio data; A reconstruction module, used for performing residual conversion on the initial audio data to reconstruct the audio data to obtain audio reconstruction data; A conversion module, used for performing digital-to-analog conversion on the audio reconstruction data to obtain an initial analog audio signal; An alignment module, used for performing feature alignment of the sorting transformation on the initial analog audio signal to obtain a target analog audio signal; The output module is used to amplify the target analog audio signal to obtain an amplified audio signal, and convert the amplified audio signal into sound waves through the earphone to output the target audio.

[0007] The third aspect of the present application provides an audio playback device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the audio playback device executes the above-mentioned audio playback method.

[0008] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned audio playback method.

[0009] In the technical solution provided by the present application, by using the generalized Gaussian mixture model (GMM) to analyze the received digital audio signal, the skewness and heavy-tail characteristics of the audio data can be captured more accurately. The accuracy of audio analysis and the quality of audio reconstruction are effectively improved. The residual conversion technology is used for audio reconstruction, which can optimize and adjust the prediction residual in the audio signal. The present application is used to restore the audio distortion caused by compression, thereby significantly improving the reconstruction quality of the audio without increasing too much computational burden. Through digital-to-analog conversion and subsequent signal amplification processing, the present method can ensure that the audio signal maintains high fidelity throughout the entire processing process. The feature alignment technology of the sorting transformation reduces the error in the conversion process and ensures the quality of the analog signal. In addition, the enhancement technology used in the signal amplification stage further ensures the dynamic range and clarity of the audio signal at the output. Not only the technical details of audio processing are considered, but also the convenience of user operation is fully considered. By integrating user interaction and feedback functions, users can directly control audio playback through headphones, such as play, pause, volume adjustment, etc., thereby improving the audio playback effect of Bluetooth headphones. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] 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 will be briefly introduced below. Obviously, the accompanying drawings described below are 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 work.

[0011] Figure 1 A schematic diagram of an embodiment of an audio playback method in an embodiment of the present application; Figure 2 Schematic diagram of an embodiment of an audio playback device in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide an audio playback method, device, equipment and storage medium. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the audio playback method in the embodiment of the present application includes: Step S101, receiving a digital audio signal through a wireless receiving module in the headset, and analyzing the digital audio signal through a generalized Gaussian mixture model to obtain audio feature data; It is understandable that the execution subject of the present application may be an audio playback device, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0014] Specifically, a digital audio signal is received through a wireless receiving module in a headset, and the digital audio signal is subjected to noise removal and normalization processing, so as to effectively eliminate background noise in the signal and unify the amplitude range of the signal, thereby obtaining a preprocessed digital signal. A generalized Gaussian mixture model is called, and its parameters are initialized to set the initial model parameters. The expectation step of the expectation maximization algorithm is used to perform expected prediction on the initial model parameters. By calculating the expected value of the preprocessed digital signal under the current model parameters, the potential variables of the signal are estimated. The maximization step of the expectation maximization algorithm is used to adjust the initial model parameters according to the estimated expected value, and the parameters are optimized to maximize the likelihood function of the data to obtain the target model parameters. The model parameters of the generalized Gaussian mixture model are updated by the target model parameters, and the model is adjusted to better fit the preprocessed digital signal to obtain an updated generalized Gaussian mixture model. The preprocessed digital signal is subjected to signal feature analysis by the updated model to generate initial feature data. The initial feature data is subjected to cluster analysis to group data with similar features to better understand and process the intrinsic structure of the signal. Each feature clustering result represents a pattern or component in the audio signal. Feature encoding is performed on the feature clustering results to convert the clustering results into audio feature data that is easier to process and apply.

[0015] Step S102: performing audio decoding and sound effect enhancement on the audio feature data to obtain initial audio data; Specifically, feature decoding is performed on the audio feature data to obtain a feature decoding sequence. Audio reconstruction is performed on the feature decoding sequence to restore the compressed and encoded information to relatively original audio signal features to obtain the first audio data. Audio noise reduction is performed on the first audio data. In the noise reduction process, a noise reduction algorithm such as adaptive filtering or spectrum subtraction is used to remove background noise and unnecessary interference in the audio signal to obtain the second audio data. Echo cancellation is performed on the second audio data. Echo cancellation uses an echo cancellation algorithm, such as an echo cancellation algorithm based on adaptive filtering, to detect and remove echo components in the audio signal to avoid interference of the echo on the audio quality, thereby obtaining the third audio data. Sound enhancement is performed on the third audio data. Sound enhancement makes the audio effect more outstanding by adjusting the frequency response, dynamic range and other audio parameters of the audio signal. It includes technologies such as equalizer adjustment, reverberation addition and dynamic range expansion, and finally obtains the fourth audio data. Spectral analysis is performed on the fourth audio data. Frequency distribution information of the audio signal is obtained by Fourier transform or other spectrum analysis methods. According to the spectrum analysis results, identify and filter out the frequency components that are important to the audio quality, perform feature screening on the feature decoding sequence, remove unnecessary frequency components, retain key features, and obtain the target feature sequence. Perform dynamic range compression on the target feature sequence. By reducing the dynamic range of the audio signal, the intensity of the audio signal is made more balanced, avoiding distortion caused by excessive volume changes, and generating compressed audio data. Perform time domain processing on the compressed audio data. Through time stretching, time compression and other technologies, adjust the distribution of the audio signal in time to make the audio playback more coherent and natural, and obtain audio data after time domain processing. Perform filtering on the audio data after time domain processing. Filtering uses high-pass filtering, low-pass filtering or band-pass filtering to filter out unnecessary frequency components and interference signals, obtain pure audio signals, and finally form the initial audio data.

[0016] Step S103, performing audio reconstruction by residual conversion on the initial audio data to obtain audio reconstruction data; Specifically, residual calculation is performed on the initial audio data. By comparing the initial audio data with its low-frequency components or compressed and encoded data, the subtle differences in the audio signal are calculated to obtain the original residual data. The reconstruction layer parameters are initialized for the original residual data. Parameter initialization includes setting the initial model parameters, including weights, biases, etc., which are used to define the initial state of the reconstruction layer. The reconstruction layer is trained for the target conversion parameters to obtain a trained reconstruction layer model. Through a large amount of training data and iterative optimization process, the parameters of the model are continuously adjusted so that it can accurately fit the original residual data. During the training process, optimization algorithms such as gradient descent are used to continuously update the model parameters according to the feedback of the loss function until the error of the model on the residual data is minimized, and the trained reconstruction layer model is obtained. The trained reconstruction layer model is used to perform residual adjustment on the original residual data. Through the prediction and correction capabilities of the reconstruction layer model, the deviations and errors in the original residual data are adjusted to generate target residual data. The initial audio data is residually compensated according to the target residual data. Combine the target residual data with the initial audio data, and compensate the residual to the initial audio data through addition or other signal fusion methods to generate compensated audio data. The purpose of residual compensation is to restore and enhance the details that were originally compressed or lost in the initial audio data. Perform reconstruction layer error correction on the compensated audio data. Through model prediction and correction, eliminate the errors and inaccuracies that may be introduced in the compensation process to obtain the corrected audio data. Smooth the corrected audio data. Smoothing eliminates sharp changes and discontinuities in the signal through filtering algorithms such as low-pass filtering and Gaussian filtering, making the audio signal more coherent and natural, and finally obtaining audio reconstruction data.

[0017] Step S104, performing digital-to-analog conversion on the audio reconstruction data to obtain an initial analog audio signal; Specifically, the audio reconstruction data is sampled, and the continuous audio signal is converted into a discrete digital signal by a sampler to obtain sampled digital audio data. The sampling frequency is selected according to the Nyquist theorem, and the sampling frequency should be at least twice the highest frequency of the audio signal to avoid the occurrence of aliasing. The sampled digital audio data is quantized, the continuous amplitude value is discretized, and it is mapped to a limited numerical range to obtain quantized audio data. The quantization accuracy, that is, the number of quantization bits, directly affects the fidelity of the audio. High-precision quantization can retain more audio details. The quantized audio data is encoded. The quantized audio data is compressed and formatted to reduce the amount of data and improve the transmission and storage efficiency to obtain encoded audio data. Common encoding technologies include pulse code modulation, adaptive differential pulse code modulation, etc. The encoded audio data is modulated, the digital signal is converted into an analog signal suitable for transmission, and the encoded audio data is converted into modulated audio data by a modulator. The selection of modulation technology depends on the transmission medium, such as amplitude modulation, frequency modulation, and phase modulation. During the transmission process, the received modulated audio data needs to be demodulated, and the modulated signal is restored to the baseband signal through the demodulator to obtain the demodulated audio data. The demodulated audio data is decoded, and the encoded data is restored to quantized audio data to obtain decoded audio data. The decoding technology corresponds to the encoding technology, such as PCM decoding, ADPCM decoding, etc. The decoded audio data is dequantized. The dequantization process restores the amplitude information of the original signal by restoring the discrete numerical value to a continuous amplitude value to obtain the dequantized audio data. The dequantized audio data is digital-to-analog converted, and the digital signal is converted into an analog signal through a digital-to-analog converter to obtain the initial analog audio signal.

[0018] Step S105, performing feature alignment of the sorting transformation on the initial analog audio signal to obtain a target analog audio signal; Specifically, data analysis is performed on the initial analog audio signal to extract multiple first signal features. These features may include frequency, amplitude, phase, etc. By analyzing these features, the basic properties and change rules of the audio signal can be understood. The multiple first signal features are sorted and transformed to be converted into an ordered signal feature sequence. According to the signal feature sequence, feature alignment parameter analysis is performed on the multiple first signal features to obtain target alignment parameters. Feature alignment parameters are the key to achieving signal feature alignment. These parameters include alignment reference points, alignment methods, and alignment ranges. Through these parameters, features at different times and frequencies are accurately aligned. According to the target alignment parameters, feature alignment is performed on multiple first signal features, and they are adjusted to a unified reference frame to obtain multiple aligned first signal features. Time domain adjustment is performed on the aligned multiple first signal features. Time domain adjustment obtains multiple second signal features by correcting the time characteristics of the signal, such as delay, jitter, etc. Frequency domain adjustment is performed on multiple second signal features, and multiple third signal features are obtained by adjusting the frequency components of the signal, such as filtering, frequency shifting, etc. Feature extraction and audio signal processing are performed on multiple third signal features to convert them into target analog audio signals. Feature extraction involves extracting the most representative features from the signal, while audio signal processing includes signal smoothing, denoising, enhancement and other processing methods to ensure that the final generated target analog audio signal has high-quality audio characteristics.

[0019] Step S106: amplify the target analog audio signal to obtain an amplified audio signal, convert the amplified audio signal into sound waves through headphones, and output the target audio.

[0020] Specifically, the target analog audio signal is pre-amplified to ensure that the signal reaches an initial amplification level suitable for further processing, thereby obtaining a pre-amplified audio signal. The amplifier parameters of the pre-amplified audio signal are initialized to determine key parameters such as the gain, input and output impedance of the amplifier to obtain the target amplification parameters. The pre-amplified audio signal is amplified according to the target amplification parameters to obtain an amplified audio signal. The amplitude of the audio signal is increased by the amplifier to reach the required volume level. The nonlinear effect in the amplification process may introduce distortion. The amplified audio signal is nonlinearly compensated by the amplifier to eliminate or reduce the nonlinear distortion generated in the amplification process to obtain a compensated amplified audio signal. The compensated amplified audio signal is distortion corrected to correct any remaining distortion caused by amplification to ensure the fidelity of the signal to obtain a corrected amplified audio signal. The corrected amplified audio signal is signal filtered to filter out unnecessary frequency components and noise to obtain a filtered amplified audio signal. After filtering, the filtered amplified audio signal is signal enhanced to improve its sound quality and clarity to obtain an enhanced amplified audio signal. The enhancement process can include adjusting the frequency response of the signal to enhance the high and low frequency parts of the audio, making the audio sound richer and more detailed. Dynamic range control is performed on the enhanced amplified audio signal to ensure that the dynamic range of the signal is within the range that the headphones can handle, avoiding signal overload and distortion. After the dynamic range control is completed, the amplified audio signal is converted into sound waves through the headphones, converting the electrical signal into sound waves and outputting the target audio.

[0021] In the embodiment of the present application, by using a generalized Gaussian mixture model (GMM) to parse the received digital audio signal, the skewness and heavy-tail characteristics of the audio data can be captured more accurately. The accuracy of audio parsing and the quality of audio reconstruction are effectively improved. The residual conversion technology is used for audio reconstruction, which can optimize and adjust the prediction residual in the audio signal. The present application is used to restore the audio distortion caused by compression, thereby significantly improving the reconstruction quality of the audio without increasing too much computational burden. Through digital-to-analog conversion and subsequent signal amplification processing, the method can ensure that the audio signal maintains high fidelity throughout the entire processing process. The feature alignment technology of the sorting transformation reduces the error in the conversion process and ensures the quality of the analog signal. In addition, the enhancement technology used in the signal amplification stage further ensures the dynamic range and clarity of the audio signal at the time of output. Not only the technical details of audio processing are considered, but also the user's convenience of operation is fully considered. By integrating user interaction and feedback functions, users can directly control audio playback through headphones, such as play, pause, volume adjustment, etc., thereby improving the audio playback effect of Bluetooth headphones.

[0022] In a specific embodiment, the process of executing step S101 may specifically include the following steps: (1) receiving a digital audio signal through a wireless receiving module in the headset, and performing noise removal and normalization processing on the digital audio signal to obtain a preprocessed digital signal; (2) Calling the generalized Gaussian mixture model and initializing the parameters of the generalized Gaussian mixture model to obtain the initial model parameters; (3) Use the expectation step of the expectation maximization algorithm to make an expected prediction of the initial model parameters and obtain the expected value; (4) Using the maximization step of the expectation maximization algorithm, the initial model parameters are processed according to the expected values ​​to obtain the target model parameters; (5) Updating the model parameters of the generalized Gaussian mixture model using the target model parameters to obtain an updated generalized Gaussian mixture model; (6) Using the updated generalized Gaussian mixture model, the preprocessed digital signal is analyzed for signal characteristics to obtain initial characteristic data; (7) Performing cluster analysis on the initial feature data to obtain multiple feature clustering results, and performing feature encoding on the multiple feature clustering results to obtain audio feature data.

[0023] Specifically, a digital audio signal is received by a wireless receiving module in the headset, and the digital audio signal is subjected to noise removal and normalization processing to obtain a preprocessed digital signal. Noise removal is achieved by a filter or other noise suppression algorithm, such as using adaptive filtering technology to eliminate environmental noise. At the same time, the normalization process adjusts the amplitude of the signal to a uniform range, making the subsequent processing steps more stable and reliable. The generalized Gaussian mixture model is called and its parameters are initialized to obtain initial model parameters. Parameter initialization generally includes setting the initial mean, variance and mixing coefficient of the model. For example, suppose there is an audio data containing speech and background music. The initial parameters of the generalized Gaussian mixture model are set by experience or using a data-driven method. For example, the initial mean of the model is set to a typical frequency component in the audio signal, the initial variance is the variation range of these frequency components, and the mixing coefficient represents the weight of each Gaussian component in the entire model. The expectation step of the expectation maximization algorithm is used to perform expected prediction on the initial model parameters to obtain expected values. The probability that each data point belongs to each Gaussian component is calculated using the current model parameters. The maximization step of the expectation maximization algorithm is adopted to process the initial model parameters according to the expected value to obtain the target model parameters. In the maximization step, the model parameters are updated by maximizing the likelihood function of the data. The mean, variance and mixing coefficient of the generalized Gaussian mixture model are updated by the probability distribution of each data point calculated by the expectation step to make these parameters more consistent with the distribution of the data. For example, by adjusting the mean of the model to make it closer to the main frequency components in the audio signal; adjusting the variance to better represent the variation range of these frequency components; adjusting the mixing coefficient to reflect the proportion of each frequency component in the overall signal. The model parameters of the generalized Gaussian mixture model are updated by the target model parameters to obtain the updated generalized Gaussian mixture model. The preprocessed digital signal is subjected to signal feature analysis by the updated generalized Gaussian mixture model to obtain the initial feature data. The purpose of signal feature analysis is to extract key features in the audio signal, such as frequency, amplitude and phase. These feature data can be classified and labeled by the generalized Gaussian mixture model. Cluster analysis is performed on the initial feature data to obtain multiple feature clustering results. By grouping similar feature data, we can better understand and process audio signals. For example, feature points with similar frequencies are clustered together to form one cluster, and feature points with similar amplitudes are clustered together to form another cluster. Multiple feature clustering results are feature encoded to obtain audio feature data. The clustering results are converted into a data format that is easier to process and analyze, such as representing each clustering result as a feature vector, or encoding it into specific frequency and amplitude values.

[0024] In a specific embodiment, the process of executing step S102 may specifically include the following steps: (1) performing feature decoding on the audio feature data to obtain a feature decoding sequence, and performing audio reconstruction on the feature decoding sequence to obtain first audio data; (2) performing audio noise reduction on the first audio data to obtain second audio data; (3) performing echo cancellation on the second audio data to obtain third audio data, and performing sound effect enhancement on the third audio data to obtain fourth audio data; (4) performing spectrum analysis on the fourth audio data to obtain a spectrum analysis result, and performing feature screening on the feature decoding sequence according to the spectrum analysis result to obtain a target feature sequence; (5) Performing dynamic range compression on the target feature sequence to obtain compressed audio data, and performing time domain processing on the compressed audio data to obtain audio data after time domain processing; (6) Filter the audio data after time domain processing to obtain initial audio data.

[0025] Specifically, feature decoding is performed on the audio feature data to convert the complex audio feature data into a more easily processable form. For example, if the audio feature data includes frequency, amplitude and phase information, these data are converted into time domain signals through feature decoding to obtain a feature decoding sequence. Audio reconstruction is performed on the feature decoding sequence, and the decoded feature data is recombined to generate a continuous audio signal to obtain the first audio data. Audio noise reduction is performed on the first audio data to obtain the second audio data. Background noise and interference in the audio signal are removed to make the audio signal clearer. Echo cancellation is performed on the second audio data to obtain the third audio data. Echo cancellation aims to eliminate echoes caused by reflections in the recording environment or the playback environment to make the audio signal clearer. An echo cancellation algorithm based on adaptive filtering is used to minimize the impact of the echo by continuously adjusting the parameters of the filter. Sound enhancement is performed on the third audio data to obtain the fourth audio data. The frequency response and dynamic range of the audio signal are adjusted to improve the sound quality and auditory effect of the audio. Equalizer adjustment, dynamic range compression and other technologies are used. For example, by adjusting the equalizer, the high-frequency and low-frequency components in the music clip are enhanced, so that the music sounds richer and more layered, and the fourth audio data is generated. The fourth audio data is subjected to spectrum analysis. The spectrum analysis obtains the spectrum distribution of the audio signal through methods such as Fourier transform or wavelet transform. The spectrum analysis result provides the amplitude and phase information of each frequency component in the audio signal. Through this information, the frequency structure of the audio signal is understood. For example, through spectrum analysis, the frequency components of the main melody and background accompaniment in the music clip are identified. According to the spectrum analysis result, the feature decoding sequence is subjected to feature screening to obtain the target feature sequence. For example, the frequency component of the main melody is retained, while the frequency component of the background noise is removed. The target feature sequence is subjected to dynamic range compression. Dynamic range compression makes the signal strength more balanced by reducing the dynamic range of the audio signal, avoids distortion caused by excessive or weak signals, and ensures that the audio signal can maintain a stable volume level during playback. The compressed audio data is subjected to time domain processing. Time domain processing includes adjusting the time characteristics of the audio signal, such as delay, time stretching or compression. Through time domain processing, the playback speed and time position of the audio signal are adjusted. The audio data after time domain processing is filtered. The filtering process uses high-pass filtering, low-pass filtering or band-pass filtering to filter out unnecessary frequency components in the audio signal, remove high-frequency noise or low-frequency interference in the signal, and obtain the initial audio data.

[0026] In a specific embodiment, the process of executing step S103 may specifically include the following steps: (1) Perform residual calculation on the initial audio data to obtain the original residual data; (2) Initialize the reconstruction layer parameters of the original residual data to obtain the target conversion parameters; (3) Perform reconstruction layer training on the target conversion parameters to obtain a trained reconstruction layer model; (4) Perform residual adjustment on the original residual data through the trained reconstruction layer model to obtain the target residual data; (5) performing residual compensation on the initial audio data according to the target residual data to obtain compensated audio data, and performing reconstruction layer error correction on the compensated audio data to obtain corrected audio data; (6) Smoothing the corrected audio data to obtain audio reconstruction data.

[0027] Specifically, the initial audio data is subjected to residual calculation, and the initial audio data is compared with its low-frequency component or a compressed and encoded version to extract high-frequency details and uncaptured detail information in the signal. The original residual data is subjected to reconstruction layer parameter initialization. The initial parameters of the reconstruction model, such as network weights and biases, are set. The initial values ​​of these parameters can be set by random initialization or based on prior knowledge. For example, when using a neural network for reconstruction layer training, a normal distribution or a uniform distribution is usually used to initialize the weights of the network to ensure that the initial state of the network has a certain diversity and robustness. The reconstruction layer is trained. The target conversion parameters are trained so that the reconstruction layer model can accurately fit the original residual data. In this process, the model is iteratively optimized using the training data set, and the weights and biases of the network are continuously adjusted through the back propagation algorithm to minimize the reconstruction error of the residual data. The original residual data is subjected to residual adjustment by the trained reconstruction layer model to obtain the target residual data. The residual adjustment process uses the trained model to predict and correct the original residual data so that it more accurately reflects the high-frequency details in the audio signal. The initial audio data is residually compensated according to the target residual data to obtain compensated audio data. Residual compensation is to add the target residual data back to the initial audio data to restore the detailed information in the signal. For example, the target residual data is added to the previously extracted low-frequency components to obtain compensated audio data containing high-frequency details. The compensated audio data is subjected to reconstruction layer error correction to obtain corrected audio data. Through model prediction and adjustment, the errors that may be introduced in the compensation process are corrected. Error correction can use the same model and method as the reconstruction layer training. Through multiple iterative optimizations, the small errors generated in the residual compensation process are eliminated to ensure the final audio signal quality. The corrected audio data is smoothed to eliminate sharp changes and discontinuities in the signal to obtain audio reconstruction data. Smoothing can use methods such as low-pass filtering and Gaussian filtering to make the audio signal more natural and smooth.

[0028] In a specific embodiment, the process of executing step S104 may specifically include the following steps: (1) Sampling the audio reconstruction data to obtain sampled digital audio data, and quantizing the sampled digital audio data to obtain quantized audio data; (2) performing data encoding on the quantized audio data to obtain encoded audio data, and performing data modulating on the encoded audio data to obtain modulated audio data; (3) demodulating the modulated audio data to obtain demodulated audio data, and decoding the demodulated audio data to obtain decoded audio data; (4) Dequantizing the decoded audio data to obtain dequantized audio data, and performing digital-to-analog conversion on the dequantized audio data to obtain an initial analog audio signal.

[0029] Specifically, the audio reconstruction data is sampled, and the continuous audio signal is converted into a discrete digital signal to obtain sampled digital audio data. The sampled digital audio data is quantized. The amplitude value of the sampling point is mapped to a limited discrete numerical range, which is usually achieved by setting the number of quantization bits. For example, the amplitude value of each sampling point is converted into a 16-bit integer to obtain quantized audio data. Quantization noise is inevitably introduced during the quantization process, but high-bit quantization can reduce this noise to a level that is difficult for the human ear to detect, thereby retaining the high fidelity of the audio signal. The quantized audio data is encoded, the audio data is compressed, and the transmission and storage efficiency is improved. Common encoding methods include pulse code modulation, adaptive differential pulse code modulation, etc. The quantized audio data is converted into a standard digital signal format through an encoder to obtain encoded audio data. The encoded audio data is modulated. Modulation is the process of converting a digital signal into an analog signal suitable for transmission, and methods such as amplitude modulation, frequency modulation, or phase modulation can be used. For example, in wireless transmission, the encoded audio data is converted into a high-frequency carrier signal through frequency modulation to generate modulated audio data. The modulated audio data is demodulated. Demodulation is the process of restoring the received modulated signal to a baseband signal, and converting the modulated audio data back to coded audio data through a demodulator. The demodulator extracts the baseband component from the carrier signal and restores the original coded audio data. After demodulation, the demodulated audio data is decoded. Decoding is the process of restoring coded data to quantized audio data, corresponding to the encoding method. For example, pulse code modulation decoding is performed on data encoded by pulse code modulation, and the coded audio data is converted to quantized audio data. The decoding process restores the discrete values ​​of the sampling points and restores the original amplitude information of the audio signal. The decoded audio data is dequantized. Dequantization is to restore the discrete quantized values ​​to continuous amplitude values, and convert the quantized audio data into an approximate continuous signal through interpolation or other reconstruction methods to obtain dequantized audio data. The dequantized audio data is digital-to-analog converted. Digital-to-analog conversion converts the digital signal into an analog signal through a digital-to-analog converter to obtain the original analog audio signal.

[0030] In a specific embodiment, the process of executing step S105 may specifically include the following steps: (1) performing data analysis on the initial analog audio signal to obtain a plurality of first signal features; (2) sorting and transforming the plurality of first signal features to obtain a signal feature sequence; (3) performing feature alignment parameter analysis on the plurality of first signal features according to the signal feature sequence to obtain target alignment parameters; (4) performing feature alignment on the plurality of first signal features according to the target alignment parameter to obtain the plurality of aligned first signal features; (5) performing time domain adjustment on the aligned multiple first signal features to obtain multiple second signal features; (6) performing frequency domain adjustment on the plurality of second signal features to obtain a plurality of third signal features; (7) Perform feature extraction and audio signal processing on the multiple third signal features to obtain a target analog audio signal.

[0031] Specifically, the initial analog audio signal is subjected to data analysis to extract multiple first signal features. These features may include frequency, amplitude, phase, etc. For example, when processing a recorded music clip, the time domain signal is converted into a frequency domain signal using Fourier transform, and the amplitude and phase information of each frequency component are extracted. The first signal features are sorted and transformed to obtain a signal feature sequence. The features are arranged according to certain rules to make the signal features more structural. For example, the frequency features are sorted according to the frequency size, or the amplitude features are sorted according to the amplitude size. According to the signal feature sequence, feature alignment parameter analysis is performed on multiple first signal features to obtain target alignment parameters. Feature alignment parameter analysis is to determine how to align these features by analyzing the patterns and rules in the feature sequence. For example, the relative time difference and phase difference between different features are calculated to obtain the time offset and phase offset parameters required for alignment. Through alignment parameter analysis, the best alignment method for each feature is determined to ensure that they are highly consistent in time and frequency. According to the target alignment parameters, multiple first signal features are feature aligned to obtain multiple aligned first signal features. Feature alignment is to adjust the signal features according to the alignment parameters so that they are aligned in time and frequency. For example, different frequency components are aligned to the same time point by time offset adjustment; and phase differences of different features are eliminated by phase adjustment to align them in phase. The aligned multiple first signal features are adjusted in time domain to obtain multiple second signal features. Time domain adjustment is to optimize the time distribution of the aligned features so that the signal is more coherent and consistent in time. For example, by time stretching or compression, the playback speed and time position of the signal are adjusted to ensure that the arrangement of each feature on the time axis is more reasonable and natural. Multiple second signal features are adjusted in frequency domain to obtain multiple third signal features. Frequency domain adjustment is to optimize the frequency distribution of the aligned features so that the signal is more balanced and clear in frequency domain. For example, by frequency translation or expansion, the frequency components of the signal are adjusted to ensure that the distribution of each feature on the frequency axis is more reasonable and orderly. By frequency domain adjustment, the frequency inconsistency in the signal is eliminated to make its frequency performance more uniform. Feature extraction and audio signal processing are performed on multiple third signal features to obtain target analog audio signals. For example, the main components in the signal are extracted by principal component analysis, or the independent components in the signal are separated by independent component analysis. Audio signal processing includes signal smoothing, denoising, enhancement and other processing methods to ensure that the final generated audio signal has high-quality audio characteristics.

[0032] In a specific embodiment, the process of executing step S106 may specifically include the following steps: (1) Pre-amplifying the target analog audio signal to obtain a pre-amplified audio signal, and initializing the amplifier parameters of the pre-amplified audio signal to obtain target amplification parameters; (2) Amplifying the target amplification parameter to obtain an amplified audio signal; (3) performing amplifier nonlinearity compensation on the amplified audio signal to obtain a compensated amplified audio signal, and performing distortion correction on the compensated amplified audio signal to obtain a corrected amplified audio signal; (4) performing signal filtering on the corrected amplified audio signal to obtain a filtered amplified audio signal, and performing signal enhancement on the filtered amplified audio signal to obtain an enhanced amplified audio signal; (5) Dynamic range control is performed on the enhanced amplified audio signal and sound wave conversion is performed on the amplified audio signal through headphones to output the target audio.

[0033] Specifically, the target analog audio signal is pre-amplified, and the audio signal is raised to a range where the amplifier can work optimally, so as to obtain a pre-amplified audio signal. The amplifier parameters are initialized for the pre-amplified audio signal to obtain an amplified audio signal. The amplifier parameter initialization includes setting key parameters such as the gain, input and output impedance of the amplifier. The gain value of the amplifier is set according to the level of the pre-amplified audio signal and the desired output volume. For example, if the level of the pre-amplified audio signal is -20dB and the desired output volume is 0dB, the gain of the amplifier should be set to 20dB. The pre-amplified audio signal is amplified according to the target amplification parameter to obtain an amplified audio signal. During the signal amplification process, the amplifier linearly amplifies the pre-amplified audio signal according to the set gain value so that its output level reaches the required standard. However, nonlinear distortion may be introduced during the amplification process. In order to eliminate these distortions, the amplified audio signal is subjected to amplifier nonlinear compensation. Nonlinear compensation corrects the nonlinear distortion introduced during the amplification process by using a feedback circuit or digital signal processing technology. For example, an adaptive filter is used to dynamically adjust the working state of the amplifier according to the difference between the input and output signals, reduce the influence of nonlinear distortion, and obtain a compensated amplified audio signal. The compensated amplified audio signal may still have some residual distortion, so distortion correction is required. Distortion correction eliminates these distortions by analyzing the harmonic and intermodulation distortion components in the signal and using the corresponding correction algorithm. For example, Fourier transform is used to analyze the spectral components in the audio signal, identify and remove the harmonic distortion introduced by the amplifier, and obtain the corrected amplified audio signal. In order to further purify the audio signal, the corrected amplified audio signal is filtered. The filtering process removes noise and unnecessary frequency components in the signal through methods such as high-pass filtering, low-pass filtering, or band-pass filtering. The filtered amplified audio signal is enhanced. Signal enhancement makes the sound quality of the audio signal better by adjusting the frequency response and dynamic range. The equalizer is used to adjust the high-frequency and low-frequency components of the audio signal so that the audio signal sounds richer and more layered. In order to ensure that the audio signal is not distorted and the volume is moderate when it is played, the enhanced amplified audio signal is subjected to dynamic range control. Dynamic range control avoids signal overload or distortion by compressing or expanding the dynamic range of the audio signal. For example, in music signal processing, dynamic range compression is used to compress the strong and weak parts of the audio signal into a certain range, making the overall volume of the music signal more balanced and moderate. The amplified audio signal is converted into sound waves through headphones to output the target audio. Sound wave conversion is the process of converting electrical signals into sound waves. Through the speaker unit of the headphones, the enhanced amplified audio signal is converted into sound waves for the user's auditory perception.

[0034] The above describes the audio playback method in the embodiment of the present application. The following describes the audio playback device in the embodiment of the present application. Figure 2 In the embodiment of the present application, an audio playback device includes: The receiving module 201 is used to receive the digital audio signal through the wireless receiving module in the headset, and analyze the digital audio signal through the generalized Gaussian mixture model to obtain audio feature data; A decoding module 202, used to perform audio decoding and sound effect enhancement on the audio feature data to obtain initial audio data; A reconstruction module 203 is used to perform residual conversion on the initial audio data to reconstruct the audio data to obtain audio reconstruction data; The conversion module 204 is used to perform digital-to-analog conversion on the audio reconstruction data to obtain an initial analog audio signal; An alignment module 205 is used to perform feature alignment of the sorting transformation on the initial analog audio signal to obtain a target analog audio signal; The output module 206 is used to amplify the target analog audio signal to obtain an amplified audio signal, and convert the amplified audio signal into sound waves through headphones to output the target audio.

[0035] Through the synergy of the above components, the received digital audio signal is parsed using a generalized Gaussian mixture model (GMM), which can more accurately capture the skewness and heavy-tail characteristics of the audio data. The accuracy of audio parsing and the quality of audio reconstruction are effectively improved. The residual conversion technology is used for audio reconstruction, which can optimize and adjust the prediction residual in the audio signal. The application is used to restore the audio distortion caused by compression, thereby significantly improving the reconstruction quality of the audio without increasing too much computational burden. Through digital-to-analog conversion and subsequent signal amplification processing, the method can ensure that the audio signal maintains high fidelity throughout the entire processing process. The feature alignment technology of the sorting transformation reduces the error in the conversion process and ensures the quality of the analog signal. In addition, the enhancement technology used in the signal amplification stage further ensures the dynamic range and clarity of the audio signal at the output. Not only the technical details of audio processing are considered, but also the user's convenience of operation is fully considered. By integrating user interaction and feedback functions, users can directly control audio playback through headphones, such as play, pause, volume adjustment, etc., thereby improving the audio playback effect of Bluetooth headphones.

[0036] The present application also provides an audio playback device, which includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor executes the steps of the audio playback method in the above embodiments.

[0037] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the audio playback method.

[0038] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0039] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0040] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An audio playback method, characterized in that: The audio playing method comprises: The digital audio signal is received through a wireless receiving module in the headset, and the digital audio signal is analyzed through a generalized Gaussian mixture model to obtain audio feature data; Performing audio decoding and sound effect enhancement on the audio feature data to obtain initial audio data; Performing audio reconstruction by residual transformation on the initial audio data to obtain audio reconstruction data; Performing digital-to-analog conversion on the audio reconstruction data to obtain an initial analog audio signal; Performing feature alignment of sorting transformation on the initial analog audio signal to obtain a target analog audio signal; The target analog audio signal is amplified to obtain an amplified audio signal, and the amplified audio signal is converted into a sound wave through the earphone to output the target audio.

2. The audio playback method according to claim 1, characterized in that: The method of receiving a digital audio signal through a wireless receiving module in the headset and parsing the digital audio signal through a generalized Gaussian mixture model to obtain audio feature data includes: Receiving a digital audio signal through a wireless receiving module in the headset, and performing noise removal and normalization processing on the digital audio signal to obtain a preprocessed digital signal; Calling a generalized Gaussian mixture model and initializing parameters of the generalized Gaussian mixture model to obtain initial model parameters; Using the expectation step of the expectation maximization algorithm to perform expected prediction on the initial model parameters to obtain expected values; Using the maximization step of the expectation maximization algorithm, the initial model parameters are processed according to the expected values ​​to obtain target model parameters; Updating the model parameters of the generalized Gaussian mixture model by using the target model parameters to obtain an updated generalized Gaussian mixture model; By using the updated generalized Gaussian mixture model, the preprocessed digital signal is subjected to signal feature analysis to obtain initial feature data; Cluster analysis is performed on the initial feature data to obtain a plurality of feature clustering results, and feature encoding is performed on the plurality of feature clustering results to obtain audio feature data.

3. The audio playback method according to claim 1, characterized in that: The step of performing audio decoding and sound effect enhancement on the audio feature data to obtain initial audio data includes: Performing feature decoding on the audio feature data to obtain a feature decoding sequence, and performing audio reconstruction on the feature decoding sequence to obtain first audio data; Performing audio noise reduction on the first audio data to obtain second audio data; performing echo cancellation on the second audio data to obtain third audio data, and performing sound effect enhancement on the third audio data to obtain fourth audio data; Performing spectrum analysis on the fourth audio data to obtain a spectrum analysis result, and performing feature screening on the feature decoding sequence according to the spectrum analysis result to obtain a target feature sequence; Performing dynamic range compression on the target feature sequence to obtain compressed audio data, and performing time domain processing on the compressed audio data to obtain audio data after time domain processing; The audio data after the time domain processing is filtered to obtain initial audio data.

4. The audio playback method according to claim 1, characterized in that: The audio reconstruction by performing residual conversion on the initial audio data to obtain audio reconstruction data includes: Performing residual calculation on the initial audio data to obtain original residual data; Initializing the reconstruction layer parameters of the original residual data to obtain target conversion parameters; Performing reconstruction layer training on the target conversion parameters to obtain a trained reconstruction layer model; Performing residual adjustment on the original residual data through the trained reconstruction layer model to obtain target residual data; Performing residual compensation on the initial audio data according to the target residual data to obtain compensated audio data, and performing reconstruction layer error correction on the compensated audio data to obtain corrected audio data; The corrected audio data is smoothed to obtain audio reconstruction data.

5. The audio playback method according to claim 1, characterized in that: The step of performing digital-to-analog conversion on the audio reconstruction data to obtain an initial analog audio signal comprises: Performing data sampling on the audio reconstruction data to obtain sampled digital audio data, and performing data quantization on the sampled digital audio data to obtain quantized audio data; Performing data encoding on the quantized audio data to obtain encoded audio data, and performing data modulation on the encoded audio data to obtain modulated audio data; Demodulating the modulated audio data to obtain demodulated audio data, and decoding the demodulated audio data to obtain decoded audio data; The decoded audio data is subjected to inverse quantization processing to obtain inverse quantized audio data, and the inverse quantized audio data is subjected to digital-to-analog conversion to obtain an initial analog audio signal.

6. The audio playback method according to claim 1, characterized in that: The step of performing feature alignment on the sorting transformation of the initial analog audio signal to obtain a target analog audio signal includes: Performing data analysis on the initial analog audio signal to obtain a plurality of first signal features; Sorting and transforming the plurality of first signal features to obtain a signal feature sequence; Performing feature alignment parameter analysis on the plurality of first signal features according to the signal feature sequence to obtain target alignment parameters; Performing feature alignment on the plurality of first signal features according to the target alignment parameter to obtain a plurality of aligned first signal features; Performing time domain adjustment on the aligned multiple first signal features to obtain multiple second signal features; Performing frequency domain adjustment on the plurality of second signal features to obtain a plurality of third signal features; Feature extraction and audio signal processing are performed on the multiple third signal features to obtain a target analog audio signal.

7. The audio playback method according to claim 1, characterized in that: The amplifying the target analog audio signal to obtain an amplified audio signal, and converting the amplified audio signal into sound waves through the earphone to output the target audio, comprises: Pre-amplifying the target analog audio signal to obtain a pre-amplified audio signal, and initializing amplifier parameters of the pre-amplified audio signal to obtain target amplification parameters; Amplifying the target amplification parameter to obtain an amplified audio signal; Performing amplifier nonlinearity compensation on the amplified audio signal to obtain a compensated amplified audio signal, and performing distortion correction on the compensated amplified audio signal to obtain a corrected amplified audio signal; Performing signal filtering on the corrected amplified audio signal to obtain a filtered amplified audio signal, and performing signal enhancement on the filtered amplified audio signal to obtain an enhanced amplified audio signal; Dynamic range control is performed on the enhanced amplified audio signal, and sound wave conversion is performed on the amplified audio signal through the earphone to output target audio.

8. An audio playback device, characterized in that: The audio playback device comprises: A receiving module, used to receive a digital audio signal through a wireless receiving module in the headset, and analyze the digital audio signal through a generalized Gaussian mixture model to obtain audio feature data; A decoding module, used to perform audio decoding and sound effect enhancement on the audio feature data to obtain initial audio data; A reconstruction module, used for performing residual conversion on the initial audio data to reconstruct the audio data to obtain audio reconstruction data; A conversion module, used for performing digital-to-analog conversion on the audio reconstruction data to obtain an initial analog audio signal; An alignment module, used for performing feature alignment of the sorting transformation on the initial analog audio signal to obtain a target analog audio signal; The output module is used to amplify the target analog audio signal to obtain an amplified audio signal, and convert the amplified audio signal into sound waves through the earphone to output the target audio.

9. An audio playback device, characterized in that: The audio playback device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instruction in the memory so that the audio playback device executes the audio playback method according to any one of claims 1 to 7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the audio playing method according to any one of claims 1 to 7 is implemented.

Images