A Bluetooth-based wireless headphone audio transmission method

Abstract

This invention discloses a Bluetooth-based wireless earphone audio transmission method. It acquires environmental audio matrix data through a sensor array in the wireless earphone; extracts noise spectrum features from the environmental audio matrix data using an improved CNN convolutional neural network; and eliminates motion friction noise in the matrix data using an adaptive filtering algorithm to obtain dual-channel noise-reduced data. The dual-channel noise-reduced data is input into a fusion model to identify the user's current scene. Combining Bluetooth RSS I signals and historical channel data, it predicts the available bandwidth within the next 5ms. Based on the scene identification result and available bandwidth, it switches the codec. Based on the codec, it segments the audio data stream into high-priority and low-priority data packets, which are transmitted through the Bluetooth main channel and auxiliary channel, respectively. This improves the audio data transmission success rate, effectively reduces audio stuttering and distortion, and enhances the user's listening experience and satisfaction.

Description

Technical Field

[0001] This invention relates to the field of wireless transmission technology, and in particular to a Bluetooth-based wireless headphone audio transmission method. Background Technology

[0002] During audio transmission, environmental noise and friction noise generated by headphone movement severely impact sound quality. Traditional noise cancellation technologies often employ filters with fixed parameters, failing to adjust noise cancellation strategies in real time according to complex and changing environments and user movement patterns. This results in poor noise cancellation performance, making it difficult for users to obtain a clear audio experience in noisy environments or during active scenarios. Furthermore, Bluetooth channel bandwidth is limited and dynamically changing. Traditional methods struggle to accurately predict available bandwidth, and when channel quality deteriorates, they cannot adjust the codec in time, easily causing audio stuttering and distortion. Moreover, existing audio data transmission methods typically do not prioritize data, failing to guarantee reliable transmission of critical audio information under bandwidth constraints, severely affecting the user's listening experience. Summary of the Invention

[0003] The purpose of this invention is to solve the above problems by designing a Bluetooth-based wireless headphone audio transmission method.

[0004] To achieve the above objectives, the technical solution of the present invention further includes the following steps in the above-mentioned Bluetooth-based wireless earphone audio transmission method:

[0005] The sensor array in the wireless earphone acquires the user's motion state, head posture, and ambient noise to obtain sensor data. The audio data and sensor data are then aligned using timestamps to obtain ambient audio matrix data.

[0006] The noise spectrum features in the environmental audio matrix data are extracted using an improved CNN convolutional neural network, and motion friction noise in the matrix data is eliminated by an adaptive filtering algorithm to obtain dual-channel noise-reduced data.

[0007] The dual-channel noise reduction data is input into the BERT-ResNet fusion model to identify the user's current scene. Combining Bluetooth RSS I signal and historical channel data, the available bandwidth within the next 5ms is predicted. The codec is switched according to the scene recognition result and the available bandwidth.

[0008] Based on the codec, the audio data stream is segmented into high-priority and low-priority data packets, which are then transmitted via the Bluetooth main channel and auxiliary channel, respectively.

[0009] Furthermore, in the aforementioned Bluetooth-based wireless earphone audio transmission method, the step of aligning audio data and sensor data using timestamps to obtain environmental audio matrix data includes:

[0010] Clock synchronization is achieved via Bluetooth protocol, and a bidirectional synchronization algorithm based on reference timestamps reduces clock deviation between sensor data and audio data to obtain synchronized data.

[0011] Establish a timestamp mapping table, and use the timestamp of the audio data as a reference to interpolate the sensor data in the synchronization data to obtain interpolated data;

[0012] The sensor state at intermediate moments is estimated using a linear interpolation method; sensor data with timestamps later than the audio frames are time-aligned to obtain aligned data.

[0013] Based on the alignment data, an environmental audio matrix data is established, which includes audio data, motion state data, head posture data, and environmental noise data. The matrix rows represent time series, and the matrix columns correspond to data dimensions.

[0014] Furthermore, in the aforementioned Bluetooth-based wireless earphone audio transmission method, the step of extracting noise spectrum features from the environmental audio matrix data using an improved CNN convolutional neural network includes:

[0015] The improved CNN convolutional neural network is based on a structure of alternating convolutional and pooling layers, including 3 convolutional layers, 2 max-pooling layers, and 2 fully connected layers. The convolutional layers use 3×3 convolutional kernels to capture local spectral features, and the pooling layers use 2×2 windows for downsampling.

[0016] Add skip connections between the inputs and outputs of adjacent convolutional layers to add the inputs to the outputs after convolution;

[0017] A batch normalization layer is added after each convolutional layer and fully connected layer to normalize the data, and the LeakyReLU activation function is used instead of the ReLU activation function of the CNN convolutional neural network.

[0018] The noise spectrum features are obtained by extracting the noise spectrum features from the environmental audio matrix data using an improved CNN convolutional neural network.

[0019] Furthermore, in the aforementioned Bluetooth-based wireless headphone audio transmission method, the step of eliminating motion friction noise in the matrix data using an adaptive filtering algorithm to obtain dual-channel noise-reduced data includes:

[0020] Acquire environmental audio matrix data, determine whether there is motion friction noise, and generate a corresponding noise reference signal;

[0021] The noise spectrum characteristics are obtained, the coefficients of the LMS filter are adjusted in real time, the adjusted filter is applied to the environmental audio matrix data, and motion friction noise is removed by an adaptive filtering algorithm to obtain dual-channel noise reduction data.

[0022] Furthermore, in the aforementioned Bluetooth-based wireless headphone audio transmission method, the step of inputting the dual-channel noise reduction data into the BERT-ResNet fusion model to identify the user's current scene includes:

[0023] Spatial features of audio data are extracted using residual blocks in the deep residual network ResNet-18, and a high-dimensional spatial feature vector is output.

[0024] The audio sequence data is processed by a pre-trained BERT model to capture the contextual dependencies of the audio signal in the time series and extract sequence features.

[0025] The spatial feature vector output by ResNet and the sequence feature vector output by BERT are concatenated to obtain a fused feature vector containing both spatial and temporal information.

[0026] The fused feature vector is input into a fully connected layer for scene classification to identify the scene in which the user is currently located.

[0027] Furthermore, in the aforementioned Bluetooth-based wireless headphone audio transmission method, the step of combining Bluetooth RSS1 signals and historical channel data to predict the available bandwidth within the next 5ms, and switching the codec based on the scene recognition result and available bandwidth, includes:

[0028] Establish a correspondence between scenarios and available bandwidth. Select high-quality codecs for quiet indoor scenarios, select codecs that balance sound quality and anti-interference capabilities for motion scenarios, and select codecs that offer low bit rate and high reliability for complex outdoor scenarios.

[0029] When the scene recognition result changes and the predicted available bandwidth exceeds the applicable range of the current codec, the codec switching process is triggered.

[0030] Furthermore, in the aforementioned Bluetooth-based wireless earphone audio transmission method, the step of segmenting the audio data stream into high-priority and low-priority data packets based on the codec, and transmitting them respectively through the Bluetooth main channel and auxiliary channel, includes:

[0031] FHSS frequency hopping spread spectrum technology is used to select frequency points with good signal quality for transmission. High-priority data packets are protected by Reed-Solomon coding for error correction. The code length is n=255 and the information bits are k=223.

[0032] Low-priority data packets utilize the ARQ (Automatic Repeat Request) mechanism. After sending the data packet, the sending end starts a timer to wait for the receiving end's acknowledgment signal.

[0033] Furthermore, in a system for implementing a Bluetooth-based wireless headphone audio transmission method, the system includes the following modules:

[0034] The matrix data acquisition module is used to acquire the user's motion state, head posture and environmental noise through the sensor group in the wireless earphone, obtain sensor data, and align the audio data and sensor data through timestamps to obtain environmental audio matrix data;

[0035] The matrix data noise reduction module is used to extract noise spectrum features from the environmental audio matrix data using an improved CNN convolutional neural network, and to eliminate motion friction noise in the matrix data through an adaptive filtering algorithm to obtain dual-channel noise-reduced data.

[0036] The decoder switching module is used to input the dual-channel noise reduction data into the BERT-ResNet fusion model to identify the user's current scene, combine the Bluetooth RSS I signal and historical channel data to predict the available bandwidth within the next 5ms, and switch the codec according to the scene recognition result and the available bandwidth.

[0037] The audio channel transmission module is used to segment the audio data stream into high-priority and low-priority data packets based on the codec, and transmit them through the Bluetooth main channel and auxiliary channel respectively.

[0038] Furthermore, in a system for implementing a Bluetooth-based wireless headphone audio transmission method, the system includes the following sub-modules:

[0039] The synchronization submodule is used for clock synchronization via Bluetooth protocol. A bidirectional synchronization algorithm based on reference timestamp reduces the clock deviation between sensor data and audio data to obtain synchronized data.

[0040] The interpolation submodule is used to establish a timestamp mapping table and interpolate the sensor data in the synchronization data based on the timestamp of the audio data to obtain interpolated data.

[0041] The alignment submodule is used to estimate the sensor state at intermediate moments using a linear interpolation method; it performs time point alignment on sensor data with timestamps later than the audio frames to obtain aligned data.

[0042] A submodule is established to create an environmental audio matrix based on the alignment data, including audio data, motion state data, head posture data, and environmental noise data. The matrix rows represent time series, and the matrix columns correspond to the data dimensions.

[0043] Furthermore, in a system for implementing a Bluetooth-based wireless headphone audio transmission method, the system includes the following sub-modules:

[0044] A submodule is configured to determine the improved CNN convolutional neural network based on a multi-layered alternating convolutional and pooling structure, including 3 convolutional layers, 2 max-pooling layers, and 2 fully connected layers. The convolutional layers use 3×3 convolutional kernels to capture local spectral features, and the pooling layers use 2×2 windows for downsampling.

[0045] Add a submodule to add skip connections between the inputs and outputs of adjacent convolutional layers, adding the input to the output after convolution;

[0046] The replacement submodule is used to add a batch normalization layer after each convolutional and fully connected layer to normalize the data, and uses the LeakyReLU activation function to replace the ReLU activation function of the CNN convolutional neural network.

[0047] The extraction submodule is used to extract noise spectrum features from the environmental audio matrix data using an improved CNN convolutional neural network, thereby obtaining noise spectrum features.

[0048] Its beneficial effects lie in acquiring user motion state, head posture, and environmental noise through a sensor group in the wireless earphone, obtaining sensor data, aligning audio data and sensor data using timestamps to obtain environmental audio matrix data; using an improved CNN convolutional neural network to extract noise spectrum features from the environmental audio matrix data, and eliminating motion friction noise in the matrix data through an adaptive filtering algorithm to obtain dual-channel noise-reduced data; inputting the dual-channel noise-reduced data into a BERT-ResNet fusion model to identify the user's current scene, combining Bluetooth RSS I signal and historical channel data to predict the available bandwidth within the next 5ms, and switching the codec based on the scene recognition result and available bandwidth; and segmenting the audio data stream into high-priority and low-priority data packets based on the codec, which are transmitted through the Bluetooth main channel and auxiliary channel respectively. 1. It can accurately extract noise spectrum features and eliminate motion friction noise, improving the noise reduction effect compared to traditional noise reduction methods, effectively improving the purity of audio, and providing users with clearer sound quality. 2. Dynamically switching codecs based on the scenario and available bandwidth improves audio transmission smoothness in complex environments, matching audio quality to the current usage scenario and channel conditions to meet users' personalized needs in different scenarios. 3. Improving the success rate of audio data transmission effectively reduces audio stuttering and distortion, enhancing the user's listening experience and satisfaction. Attached Figure Description

[0049] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0050] Figure 1 This is a schematic diagram of the first embodiment of a Bluetooth-based wireless headphone audio transmission method according to the present invention;

[0051] Figure 2 This is a schematic diagram of a second embodiment of a Bluetooth-based wireless headphone audio transmission method according to an embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram of a third embodiment of a Bluetooth-based wireless headphone audio transmission method according to the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0054] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0055] The present invention will now be described in detail with reference to the accompanying drawings, such as... Figure 1 As shown, a Bluetooth-based wireless headphone audio transmission method includes the following steps:

[0056] Step 101: Obtain the user's motion state, head posture, and ambient noise through the sensor group in the wireless earphone to obtain sensor data. Align the audio data and sensor data with timestamps to obtain ambient audio matrix data.

[0057] Specifically, in this embodiment, clock synchronization is performed via Bluetooth protocol, and a bidirectional synchronization algorithm based on reference timestamps is used to reduce clock deviation between sensor data and audio data to obtain synchronized data.

[0058] Establish a timestamp mapping table, and use the timestamp of the audio data as a reference to interpolate the sensor data in the synchronization data to obtain interpolated data;

[0059] The sensor state at intermediate moments is estimated using a linear interpolation method; sensor data with timestamps later than the audio frames are time-aligned to obtain aligned data.

[0060] An environmental audio matrix is ​​built based on aligned data, including audio data, motion state data, head posture data, and environmental noise data. The matrix rows represent time series data, and the matrix columns correspond to the data dimensions.

[0061] Step 102: Extract noise spectrum features from the environmental audio matrix data using an improved CNN convolutional neural network, and eliminate motion friction noise in the matrix data using an adaptive filtering algorithm to obtain dual-channel noise-reduced data;

[0062] Specifically, the improved CNN convolutional neural network in this embodiment is based on a structure of alternating convolutional and pooling layers, including 3 convolutional layers, 2 max-pooling layers, and 2 fully connected layers. The convolutional layers use 3×3 convolutional kernels to capture local spectral features, and the pooling layers use 2×2 windows for downsampling;

[0063] Add skip connections between the inputs and outputs of adjacent convolutional layers to add the inputs to the outputs after convolution;

[0064] A batch normalization layer is added after each convolutional layer and fully connected layer to normalize the data, and the LeakyReLU activation function is used instead of the ReLU activation function of the CNN convolutional neural network.

[0065] An improved CNN convolutional neural network is used to extract noise spectrum features from environmental audio matrix data, thus obtaining noise spectrum features.

[0066] Acquire environmental audio matrix data, determine whether there is motion friction noise, and generate a corresponding noise reference signal;

[0067] The noise spectrum characteristics are obtained, the coefficients of the LMS filter are adjusted in real time, the adjusted filter is applied to the environmental audio matrix data, and motion friction noise is removed by an adaptive filtering algorithm to obtain dual-channel noise reduction data.

[0068] Step 103: Input the dual-channel noise reduction data into the BERT-ResNet fusion model to identify the user's current scene. Combine the Bluetooth RSSI signal and historical channel data to predict the available bandwidth within the next 5ms. Switch the codec based on the scene recognition result and the available bandwidth.

[0069] Specifically, in this embodiment, the residual blocks in the deep residual network ResNet-18 are used to extract the spatial features of the audio data and output a high-dimensional spatial feature vector.

[0070] The audio sequence data is processed by a pre-trained BERT model to capture the contextual dependencies of the audio signal in the time series and extract sequence features.

[0071] The spatial feature vector output by ResNet and the sequence feature vector output by BERT are concatenated to obtain a fused feature vector containing both spatial and temporal information.

[0072] The fused feature vector is input into a fully connected layer for scene classification, which identifies the scene in which the user is currently located.

[0073] Establish a correspondence between scenarios and available bandwidth. Select high-quality codecs for quiet indoor scenarios, select codecs that balance sound quality and anti-interference capabilities for motion scenarios, and select codecs that offer low bit rate and high reliability for complex outdoor scenarios.

[0074] When the scene recognition result changes and the predicted available bandwidth exceeds the applicable range of the current codec, the codec switching process is triggered.

[0075] Step 104: Based on the codec, the audio data stream is segmented into high-priority and low-priority data packets, which are then transmitted through the Bluetooth main channel and auxiliary channel, respectively.

[0076] Specifically, in this embodiment, FHSS frequency hopping spread spectrum technology is used to select frequency points with good signal quality for transmission. High-priority data packets are protected by Reed-Sol omon coding for error correction. The code length is n=255 and the information bits are k=223.

[0077] Low-priority data packets utilize the ARQ (Automatic Repeat Request) mechanism. After sending the data packet, the sending end starts a timer to wait for the receiving end's acknowledgment signal.

[0078] Its beneficial effects lie in acquiring user motion state, head posture, and environmental noise through a sensor array in the wireless earphone, obtaining sensor data, aligning audio data and sensor data using timestamps to obtain environmental audio matrix data; using an improved CNN convolutional neural network to extract noise spectrum features from the environmental audio matrix data, and eliminating motion friction noise in the matrix data through an adaptive filtering algorithm to obtain dual-channel noise-reduced data; inputting the dual-channel noise-reduced data into a BERT-ResNet fusion model to identify the user's current scene, combining Bluetooth RSS I signal and historical channel data to predict the available bandwidth within the next 5ms, and switching the codec based on the scene recognition result and available bandwidth; and segmenting the audio data stream into high-priority and low-priority data packets based on the codec, which are transmitted through the Bluetooth main channel and auxiliary channel respectively. 1. It can accurately extract noise spectrum features and eliminate motion friction noise, improving the noise reduction effect compared to traditional noise reduction methods, effectively improving the purity of audio, and providing users with clearer sound quality. 2. Dynamically switching codecs based on the scenario and available bandwidth improves audio transmission smoothness in complex environments, matching audio quality to the current usage scenario and channel conditions to meet users' personalized needs in different scenarios. 3. Improving the success rate of audio data transmission effectively reduces audio stuttering and distortion, enhancing the user's listening experience and satisfaction.

[0079] Please see Figure 2 In a Bluetooth-based wireless earphone audio transmission method, obtaining environmental audio matrix data by aligning audio data and sensor data using timestamps includes the following steps:

[0080] Step 201: Perform clock synchronization via Bluetooth protocol. A bidirectional synchronization algorithm based on reference timestamps reduces the clock deviation between sensor data and audio data to obtain synchronized data.

[0081] Step 202: Establish a timestamp mapping table, and use the timestamp of the audio data as a reference to interpolate the sensor data in the synchronization data to obtain interpolated data;

[0082] Step 203: Estimate the sensor state at intermediate moments using linear interpolation; perform time point alignment on sensor data with timestamps later than the audio frames to obtain aligned data;

[0083] Step 204: Based on the alignment data, establish an environmental audio matrix data including audio data, motion state data, head posture data, and environmental noise data. The matrix rows represent time series, and the matrix columns correspond to the data dimensions.

[0084] Please see Figure 3In a Bluetooth-based wireless headphone audio transmission method, inputting dual-channel noise-reduced data into a BERT-ResNet fusion model to identify the user's current scene includes the following steps:

[0085] Step 301: Extract spatial features of audio data using residual blocks in the deep residual network ResNet-18, and output a high-dimensional spatial feature vector;

[0086] Step 302: Process the audio sequence data using a pre-trained BERT model to capture the contextual dependencies of the audio signal in the time series and extract sequence features;

[0087] Step 303: Concatenate the spatial feature vector output by ResNet and the sequence feature vector output by BERT to obtain a fused feature vector containing spatial and temporal information;

[0088] Step 304: Input the fused feature vector into the fully connected layer for scene classification to identify the scene in which the user is currently located.

[0089] Specifically, the present invention can also be implemented through the following steps:

[0090] I. Data Acquisition and Alignment

[0091] (I) Sensor Assembly Composition and Data Acquisition

[0092] The built-in sensor array in the wireless earbuds consists of multiple types of sensors working together to comprehensively acquire user status and environmental information. Specifically, these include:

[0093] Motion and attitude sensors: These include a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer. The accelerometer measures the user's acceleration in three orthogonal directions, determining whether the user is stationary, walking, running, jumping, or in other motion states. The gyroscope senses the head's angular velocity along the three axes in real time. Combined with the orientation information provided by the magnetometer, it accurately determines the head's pitch, yaw, and roll attitudes, such as vertical, horizontal, and lateral head movements.

[0094] Ambient noise acquisition sensor: Employs a multi-microphone array, typically positioned on the surface of the earphone shell, to acquire ambient noise from different angles and positions, ensuring comprehensive capture of surrounding sound signals such as traffic noise, human voices, and wind noise.

[0095] Each sensor acquires data at a uniform sampling frequency of 44.1kHz, which is consistent with common audio sampling frequencies, facilitating subsequent data processing and synchronization.

[0096] (ii) Timestamp Alignment Mechanism

[0097] To achieve accurate synchronization between audio data and sensor data, the following timestamp alignment method is used:

[0098] Timestamp generation: During the acquisition of audio and sensor data, the high-precision clock module built into the headphones generates a timestamp for each frame of data. The timestamp is accurate to the microsecond level, ensuring the accuracy of the time stamp.

[0099] Clock synchronization: During the connection initialization phase, the headphones and paired Bluetooth devices (phones, players, etc.) synchronize their clocks via the Bluetooth protocol. A bidirectional synchronization algorithm based on reference timestamps is used to reduce clock deviation between the two devices and ensure the consistency of the entire system time.

[0100] Data alignment processing: A timestamp mapping table is established, and the sensor data is interpolated or resampled based on the timestamp of the audio data. For sensor data with timestamps earlier than the current audio frame timestamp, a linear interpolation method is used to estimate the sensor state at intermediate moments; for sensor data with timestamps later than the audio frame, alignment is performed at a suitable time point, ultimately forming an environmental audio matrix data containing audio data, motion state data, head posture data, and environmental noise data. The rows of the matrix represent time series, and the columns correspond to different types of data dimensions.

[0101] II. Noise Elimination Treatment

[0102] (I) Improved CNN Convolutional Neural Network

[0103] An improved CNN convolutional neural network was designed for noise spectrum feature extraction from environmental audio matrix data. Its network structure and improvements are as follows:

[0104] Network architecture: It adopts a structure of alternating convolutional and pooling layers, specifically including 3 convolutional layers, 2 max pooling layers, and 2 fully connected layers. The convolutional layers use 3×3 convolutional kernels to capture local spectral features; the pooling layers use 2×2 windows for downsampling to reduce data dimensionality.

[0105] Improvement measures:

[0106] Residual connections: Residual connections are introduced between adjacent convolutional layers, that is, skip connections are added between the input and the output, and the input is directly added to the output after convolution. This effectively solves the gradient vanishing problem in deep network training and enables the network to learn more complex noise spectrum features.

[0107] Batch normalization: Add a batch normalization layer after each convolutional layer and fully connected layer to normalize the data, accelerate network convergence, and improve the training stability and generalization ability of the model.

[0108] Activation function: The LeakyReLU activation function is used instead of the traditional ReLU function to avoid the "death" of neurons in the negative region and enhance the nonlinear expressive ability of the network.

[0109] (II) Adaptive Filtering Algorithm

[0110] Based on the extracted noise spectrum features, an adaptive least mean square (LMS) filtering algorithm is used to eliminate motion friction noise. The specific steps are as follows:

[0111] Noise reference signal generation: Using the ambient noise data collected by the microphone array, combined with the motion state and head posture information obtained by the sensor group, it is determined whether there is motion friction noise (noise generated by the friction between the headphones and the head) and a corresponding noise reference signal is generated.

[0112] Filter coefficient adjustment: The coefficients of the LMS filter are adjusted in real time based on the noise spectrum features extracted by the improved CNN. The filter adopts a transverse FIR structure. Noise cancellation: The adjusted filter is applied to the environmental audio matrix data, and motion friction noise is removed through filtering to obtain dual-channel (left channel and right channel) noise-reduced data.

[0113] III. Scene Recognition and Codec Switching

[0114] (I) BERT-ResNet Fusion Model

[0115] To accurately identify the user's current scenario, a BERT-ResNet fusion model was constructed. The fusion process is as follows:

[0116] Data preprocessing: The dual-channel denoised data is converted into a format suitable for both network inputs. For the ResNet part, the audio data of each channel is treated as a two-dimensional matrix (time-frequency domain) and used as the input to ResNet; for the BERT part, the audio data is converted into a sequence, and the audio feature at each time point is used as a token and input into BERT.

[0117] Feature extraction:

[0118] The ResNet part employs a deep residual network, ResNet-18, which extracts spatial features (spatial distribution of the spectrum, energy concentration regions, etc.) from audio data through multiple residual blocks, and outputs a high-dimensional spatial feature vector.

[0119] BERT part: The pre-trained BERT model is used to process audio sequence data, capture the contextual dependencies of audio signals in the time series, and extract sequence features.

[0120] Feature fusion: The spatial feature vector output by ResNet and the sequence feature vector output by BERT are concatenated to form a fused feature vector containing spatial and temporal information. This fused feature vector is then input into a fully connected layer for scene classification, identifying the user's current scene, such as indoor, outdoor, sports, or quiet.

[0121] (II) Available Bandwidth Prediction

[0122] The available bandwidth within the next 5ms is predicted by combining Bluetooth RSSI signal and historical channel data. The specific method is as follows:

[0123] Bluetooth RSSI signal acquisition: The received signal strength indicator (RSSI) of the Bluetooth signal is acquired in real time at a frequency of 100Hz, reflecting the signal strength and channel quality of the current Bluetooth link.

[0124] Historical channel data management: Establish a historical channel database to store channel parameters such as RSSI value, transmission rate, and bit error rate within a past period (1 minute), and index them in chronological order.

[0125] Prediction Model: A Long Short-Term Memory (LSTM) network is used as the prediction model. Inputs are the current RSSI value and historical channel data, and the output is the available bandwidth within the next 5ms. The LSTM model can capture the time-series correlation of channel parameters and adapt to dynamic changes in the channel.

[0126] (III) Codec Switching Strategy

[0127] Based on the scene recognition results and available bandwidth, different encoding and decoding schemes are pre-set, and the specific switching rules are as follows:

[0128] Scene-Bandwidth Mapping Table: Establishes the correspondence between scenes and available bandwidth, for example:

[0129] For quiet indoor scenarios: where available bandwidth is high (>2Mbps), select a high-quality audio codec (LDAC);

[0130] For motion scenarios: with moderate available bandwidth (1-2Mbps), choose a codec that balances sound quality and interference resistance (aptX Adaptive);

[0131] For complex outdoor scenarios with low available bandwidth (<1Mbps), choose a low bit rate, high reliability codec (SBC).

[0132] Real-time switching mechanism: When the scene recognition result changes or the predicted available bandwidth exceeds the applicable range of the current codec, the codec switching process is triggered. During the switching process, the current audio transmission is stopped first, and the codec is reconfigured and negotiated to ensure the stability and compatibility of the switching.

[0133] IV. Data Fragmentation and Transmission

[0134] (I) Data Packet Priority Division

[0135] Based on the importance of the audio data and the requirements for real-time performance, the audio data stream is divided into high-priority and low-priority data packets:

[0136] High-priority data packets contain crucial audio information, such as the fundamental frequency and formant parameters of the speech, which are essential for audio intelligibility and quality. For music signals, high-priority data packets include the main melody and harmonic information in the low-frequency range.

[0137] Low-priority data packets: These include minor audio information, such as high-frequency detail effects and ambient reverberation. These data can improve audio quality to some extent, but can be sacrificed when bandwidth is limited.

[0138] (II) Channel Allocation and Transmission Strategy

[0139] Channel division: The Bluetooth link is divided into a primary channel and an auxiliary channel. The primary channel is used to transmit high-priority data packets and adopts frequency hopping spread spectrum (FHSS) technology to select the frequency point with better signal quality for transmission to ensure the reliability of transmission. The auxiliary channel is used to transmit low-priority data packets and adopts fixed frequency point transmission to improve spectrum utilization.

[0140] Encoding and retransmission mechanisms:

[0141] High-priority data packets: Reed-Solomon (RS) encoding is used for error correction protection. An RS code with a code length of n=255 and information bits of k=223 is selected, capable of correcting 16 symbol errors. During transmission, RS-encoded redundancy information is added to each high-priority data packet, and the receiving end recovers potentially erroneous data through RS decoding.

[0142] Low-priority data packets: Automatic Repeat Request (ARQ) mechanism is used. After sending a data packet, the sender starts a timer to wait for an acknowledgment signal from the receiver. If no acknowledgment is received within the specified time, the data packet is retransmitted. The number of retransmissions is set to 3 to avoid increased latency due to frequent retransmissions.

[0143] (III) Fragmentation Strategy

[0144] Determine the appropriate segment size based on the codec's output data rate and available bandwidth. Set the transmission interval to 1ms to ensure real-time segmentation. Divide the audio data stream into multiple data packets according to the segment size, and label each data packet with priority and sequence number to facilitate reassembly and processing at the receiving end.

[0145] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A Bluetooth-based wireless earphone audio transmission method, characterized in that, The wireless earphone audio transmission method comprises the following steps: Obtain the user motion state, head posture and environmental noise through the sensor group in the wireless earphone to obtain sensor data, align the audio data and the sensor data through a timestamp to obtain environmental audio matrix data; Extract the noise spectrum features in the environmental audio matrix data by using an improved CNN convolutional neural network, eliminate the motion friction noise in the matrix data by using an adaptive filtering algorithm to obtain double-channel noise reduction data; Input the double-channel noise reduction data into a BERT-ResNet fusion model to identify the current scene of the user, combine the Bluetooth RSSI signal and the historical channel data to predict the available bandwidth within the next 5ms, and switch the codec according to the scene identification result and the available width; Based on the codec, the audio data stream is segmented into high-priority and low-priority data packets, which are transmitted through the main channel and the auxiliary channel of Bluetooth, respectively.

2. The Bluetooth-based wireless earphone audio transmission method of claim 1, wherein, The environmental audio matrix data is obtained by aligning the audio data and the sensor data through a timestamp, comprising: Synchronize the clock through the Bluetooth protocol, reduce the clock deviation of the sensor data and the audio data based on the bidirectional synchronization algorithm of the reference timestamp to obtain synchronization data; Establish a timestamp mapping table, and perform interpolation processing on the sensor data in the synchronization data based on the timestamp of the audio data to obtain interpolation data; Estimate the sensor state at the intermediate time by using the linear interpolation method; align the time points of the sensor data with the timestamp later than the audio frame to obtain alignment data; Based on the alignment data, the environmental audio matrix data including the audio data, the motion state data, the head posture data and the environmental noise data is established, the matrix row represents the time sequence, and the matrix column corresponds to the data dimension.

3. The Bluetooth-based wireless earphone audio transmission method of claim 1, wherein, The improved CNN convolutional neural network is based on the structure of multiple convolutional layers and pooling layers alternately, comprising three convolutional layers, two maximum pooling layers and two fully connected layers, the convolutional layer uses a 3*3 convolutional kernel to capture local spectrum features, and the pooling layer uses a 2*2 window for downsampling; Add a skip connection between the input and output of adjacent convolutional layers to add the input to the output after convolution processing; Add a batch normalization layer after each convolutional layer and fully connected layer to normalize the data, and use a LeakyReLU activation function to replace the ReLU activation function of the CNN convolutional neural network; The improved CNN convolutional neural network extracts the noise spectrum features in the environmental audio matrix data to obtain the noise spectrum features. The double-channel noise reduction data is obtained by eliminating the motion friction noise in the matrix data through an adaptive filtering algorithm, comprising:

4. The Bluetooth-based wireless earphone audio transmission method of claim 2, wherein, Obtain the environmental audio matrix data, determine whether there is motion friction noise at present and generate the corresponding noise reference signal; Obtain the noise spectrum features, adjust the coefficients of the LMS filter in real time, apply the adjusted filter to the environmental audio matrix data, remove the motion friction noise through the adaptive filtering algorithm, and obtain the double-channel noise reduction data. ​ 5. The Bluetooth-based wireless earphone audio transmission method of claim 1, wherein, The inputting of the double-channel noise reduction data into the BERT-ResNet fusion model identifies the current scene of the user, and comprises the following steps: The residual block in the deep residual network ResNet-18 is used to extract the spatial features of the audio data, and a high-dimensional spatial feature vector is outputted; The pre-trained BERT model is used to process the audio sequence data, capture the context dependence relationship of the audio signal in the time sequence, and extract the sequence features; The spatial feature vector outputted by the ResNet and the sequence feature vector outputted by the BERT are spliced to obtain a fusion feature vector containing spatial and time information; The fusion feature vector is inputted into a full connection layer for scene classification to identify the current scene of the user.

6. The Bluetooth-based wireless earphone audio transmission method of claim 1, wherein, The combination of the Bluetooth RSSI signal and the historical channel data is used to predict the available bandwidth within the next 5ms, and the codec is switched according to the scene recognition result and the available width, and comprises the following steps: A corresponding relationship between the scene and the available bandwidth is established, a high-audio-quality codec is selected for a quiet indoor scene, a codec considering both audio quality and anti-interference capability is selected for a sports scene, and a low-code-rate and reliable codec is selected for a complex outdoor scene; When the scene recognition result changes and the predicted available bandwidth exceeds the applicable range of the current codec, a codec switching process is triggered.

7. The Bluetooth-based wireless earphone audio transmission method of claim 1, wherein, The audio data stream is divided into high-priority and low-priority data packets based on the codec, and is transmitted through the Bluetooth main channel and the auxiliary channel respectively, and comprises the following steps: The FHSS frequency hopping spread spectrum technology is used to select a frequency point with good signal quality for transmission, the high-priority data packet is protected by error correction using Reed-Solomon coding, and the RS code with a code length of n=255 and an information bit of k=223 is used; The low-priority data packet uses the ARQ automatic repeat request mechanism, and the sending end starts a timer to wait for the confirmation signal of the receiving end after sending the data packet.

8. A system for implementing the Bluetooth based wireless earphone audio transmission method as claimed in claim 1, wherein, The system comprises the following modules: A matrix data acquisition module is configured to acquire user motion state, head posture and environmental noise through a sensor group in a wireless earphone, obtain sensor data, align audio data and sensor data through a time stamp, and obtain environmental audio matrix data; A matrix data noise reduction module is configured to extract noise spectrum features in the environmental audio matrix data using an improved CNN convolutional neural network, eliminate motion friction noise in the matrix data through an adaptive filtering algorithm, and obtain double-channel noise reduction data; A decoder switching module is configured to input the double-channel noise reduction data into a BERT-ResNet fusion model to identify the current scene of the user, combine a Bluetooth RSSI signal and historical channel data to predict the available bandwidth within the next 5ms, and switch the codec according to the scene recognition result and the available width; An audio channel transmission module is configured to divide the audio data stream into high-priority and low-priority data packets based on the codec, and transmit the data packets through the Bluetooth main channel and the auxiliary channel respectively.

9. A system for implementing a Bluetooth based wireless earphone audio transmission method as claimed in claim 1, characterized in that, The matrix data acquisition module comprises the following sub-modules: A synchronization submodule is configured to synchronize clocks through a Bluetooth protocol, reduce clock deviation of sensor data and audio data based on a reference timestamp-based two-way synchronization algorithm, and obtain synchronized data. An interpolation submodule is configured to establish a timestamp mapping table, perform interpolation processing on the sensor data in the synchronized data based on the timestamp of the audio data, and obtain interpolated data. An alignment submodule is configured to estimate the sensor state at an intermediate time point using a linear interpolation method. Sensor data with a timestamp later than an audio frame is time-point aligned to obtain aligned data. An establishment submodule is configured to establish environmental audio matrix data including audio data, motion state data, head pose data, and environmental noise data based on the aligned data, wherein a matrix row represents a time sequence and a matrix column corresponds to a data dimension.

10. A system for implementing the Bluetooth-based wireless earphone audio transmission method according to claim 1, characterized in that, The system comprises the following modules: A setting submodule is configured to determine that an improved CNN convolutional neural network is based on an alternating structure of multiple convolutional layers and pooling layers, including three convolutional layers, two maximum pooling layers, and two fully connected layers, a convolutional layer uses a 3*3 convolutional kernel to capture local spectral features, and a pooling layer uses a 2*2 window for down-sampling. An adding submodule is configured to add a skip connection between the input and output of adjacent convolutional layers, and add the input to the output after convolution processing. A replacement submodule is configured to add a batch normalization layer after each convolutional layer and fully connected layer, normalize the data, and replace the ReLU activation function of the CNN convolutional neural network with a LeakyReLU activation function. An extraction submodule is configured to extract noise spectral features in the environmental audio matrix data using the improved CNN convolutional neural network, and obtain noise spectral features.

Images