Bluetooth earphone signal interference suppression system based on graph neural network

Abstract

The invention discloses a Bluetooth earphone signal interference suppression system based on a graph neural network, and the system comprises a data collection and processing module which is used for collecting operation data and carrying out the normalization processing; the interference topology construction module is used for constructing an interference topological graph according to the operation data; the graph neural network inference module is used for generating an interference characterization vector and an interference correlation characterization vector; the priori generation module is used for constructing a priori vector and generating a corresponding hyper-parameter set; the frequency hopping consistency construction module is used for constructing a frequency hopping consistency constraint; the signal reconstruction module is used for executing sparse Bayesian compressed sensing to solve and reconstruct a Bluetooth signal; the communication parameter generation module is used for generating a frequency hopping sequence adjustment parameter, a channel selection parameter, a transmitting power parameter and a coding parameter; and the link adjusting module is used for adjusting the wireless communication link of the Bluetooth earphone in real time and maintaining stable transmission. According to the invention, interference is suppressed by using the graph neural network and sparse compressed sensing, and the Bluetooth signal reconstruction and transmission stability is improved.

Description

Technical Field

[0001] This invention relates to the field of radio frequency interference suppression technology, and in particular to a Bluetooth headset signal interference suppression system based on graph neural networks. Background Technology

[0002] In frequency-hopping communication of short-range wireless devices such as Bluetooth headsets, frequency band sharing, dense device density, and multi-source radio frequency interference make wireless links highly susceptible to signal impact. Existing Bluetooth anti-interference technologies mainly rely on adaptive frequency-hopping algorithms with fixed thresholds, avoidance strategies based on power detection, or simple channel quality assessment methods. These methods make decisions based only on local signal characteristics and cannot characterize the topological relationships between multiple devices, multiple frequency bands, and multiple time-series interferences. They also cannot identify the propagation and coupling paths of interference from a global perspective, making it difficult to cope with high-intensity, unstable, and multi-source superimposed radio frequency interference environments.

[0003] Meanwhile, traditional signal recovery methods lack modeling of interference structures in Bluetooth signal reconstruction, often assuming that interference is uncorrelated across frequency bands. This makes it difficult for compressed sensing-based reconstruction techniques to utilize structured information between frequency bands, resulting in discontinuous sparse coefficients during frequency hopping cycles and low reconstruction accuracy. Furthermore, existing link conditioning strategies typically rely on simple rules and cannot combine signal reconstruction results with interference trends for dynamic decision-making, thus failing to achieve stable Bluetooth communication links under complex interference conditions.

[0004] Therefore, how to provide a Bluetooth headset signal interference suppression system based on graph neural networks is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a Bluetooth headset signal interference suppression system based on graph neural networks. This invention utilizes graph neural networks and sparse compressed sensing to suppress interference and improve the stability of Bluetooth signal reconstruction and transmission.

[0006] A Bluetooth headset signal interference suppression system based on a graph neural network according to an embodiment of the present invention includes:

[0007] The data acquisition and processing module is used to acquire wireless channel operation data of frequency hopping communication and perform normalization processing;

[0008] The interference topology construction module constructs an interference topology graph containing node attribute vectors and edge attribute vectors based on the running data.

[0009] The graph neural network inference module is used to generate interference representation vectors and interference-related representation vectors, and to perform trend inference to generate interference trend prediction results.

[0010] The prior generation module is used to construct sparse prior vectors and correlated prior vectors and generate corresponding hyperparameter sets.

[0011] The frequency hopping consistency construction module is used to establish frequency hopping mapping relationships and construct frequency hopping consistency constraints;

[0012] The signal reconstruction module performs sparse Bayesian compressed sensing to solve and reconstruct the Bluetooth signal based on prior, hyperparameter and frequency hopping consistency constraints;

[0013] The communication parameter generation module generates frequency hopping sequence adjustment parameters, channel selection parameters, transmit power parameters, and coding parameters based on the reconstructed signal quality indicators and interference trend prediction results.

[0014] The link adjustment module adjusts the Bluetooth headset's wireless communication link in real time and maintains stable transmission.

[0015] Optionally, modules can be integrated using the following methods:

[0016] Collect wireless channel operation data of Bluetooth headsets during frequency hopping communication, and perform normalization processing on the wireless channel operation data according to frequency index and time order;

[0017] An interference topology map is constructed based on the normalized wireless channel operation data;

[0018] The interference topology map is input into the graph neural network to obtain the interference representation vector and interference correlation representation vector of the frequency hopping channel. Trend inference is then performed to generate interference trend prediction results.

[0019] The interference representation vector and the interference correlation representation vector are constructed into sparse prior vectors and correlation prior vectors for sparse Bayesian compressed sensing, and a set of sparsity hyperparameters and a set of correlation hyperparameters are established.

[0020] Establish a frequency hopping mapping relationship based on the frequency hopping sequence, obtain the sparse coefficients of the current frequency hopping period, and construct frequency hopping consistency constraints;

[0021] Perform sparse Bayesian compressed sensing to estimate sparse coefficients and reconstruct the Bluetooth signal affected by interference, and obtain the reconstructed Bluetooth signal and signal quality indicators.

[0022] Based on the reconstructed signal quality indicators and interference trend prediction results, calculate the frequency hopping sequence adjustment parameters, channel selection parameters, transmit power parameters, and coding parameters;

[0023] The wireless communication link of the Bluetooth headset is adjusted in real time to maintain stable signal transmission of the Bluetooth headset under multi-source interference conditions.

[0024] Optionally, the construction of the interference topology map includes:

[0025] The normalized wireless channel operation data is arranged according to the frequency hopping channel index order, and each frequency hopping channel and each identifiable interference source is recorded as a node to form a node set.

[0026] Based on the normalized received signal strength data, channel occupancy data, signal-to-noise ratio data, neighboring device frequency band information data, and multi-source interference energy distribution data, a node attribute vector is constructed for each node in the node set, forming a node attribute vector set.

[0027] Interference coupling degree is generated based on the relationship between the received signal strength change, the channel occupancy rate change, and the interference energy change between any two nodes in the node attribute vector set. The interference coupling degree is then combined in a fixed order to construct an edge attribute vector, forming an edge attribute vector set.

[0028] Establish the connection relationship between nodes based on the node set and the edge attribute vector set, and establish an edge between each node and the corresponding node determined by the edge attribute vector to form an edge set;

[0029] A graph structure is constructed using the set of nodes, the set of edges, the set of node attribute vectors, and the set of edge attribute vectors, and this graph structure is labeled as an interference topology graph.

[0030] Optionally, the generation of the interference characterization vector, the interference correlation characterization vector, and the interference trend prediction result includes:

[0031] The interference topology graph is input into the graph neural network. The set of nodes, the set of edges, the set of node attribute vectors, and the set of edge attribute vectors in the interference topology graph are input into the graph structure receiver of the graph neural network in a fixed order.

[0032] The node attribute vector of each node in the node attribute vector set is used as the initial node feature of that node, and the edge attribute vector of each edge in the edge attribute vector set is used as the initial edge feature of that edge, thus forming the initial node feature set and the initial edge feature set.

[0033] Based on the initial node feature set, a neighborhood aggregation operation is performed on each node at the neighborhood aggregation operation end of the graph neural network to form a set of node aggregation vectors;

[0034] The node aggregation vector corresponding to each node in the node aggregation vector set is fused with the attribute vectors of all edges connected to that node in the initial edge feature set in a fixed manner to form a node update vector set.

[0035] Replace the node attribute vector of each node in the node update vector set with the node update vector corresponding to that node to form the node attribute vector set after node update;

[0036] Record the node update vector corresponding to each node in the set of updated node attribute vectors as the interference characterization vector of the frequency hopping channel corresponding to that node, and form an interference characterization vector set.

[0037] Based on the updated set of node attribute vectors and the initial set of edge features, the edge attribute vector of each edge and the node update vectors of the two connected nodes are combined in a fixed order to form a set of interference association representation vectors.

[0038] Arrange the set of interference characterization vectors in order of frequency hopping cycle, record the interference characterization vectors corresponding to each frequency hopping cycle, and form an interference characterization sequence.

[0039] Based on the changing relationship of the interference characterization vector between each frequency hopping cycle in the interference characterization sequence, trend inference is performed, and the trend inference results are recorded in the order of frequency hopping cycles to form the interference trend prediction results.

[0040] Optionally, the establishment of the sparsity hyperparameter set and the correlation hyperparameter set includes:

[0041] The interference representation vectors are recorded in a fixed order, and each interference representation vector is used as the basic data for constructing sparse Bayesian compressed sensing sparse prior vectors. The sparse prior vectors are arranged in the order of their corresponding nodes to form a sparse prior vector set.

[0042] The interference association representation vectors are recorded in a fixed order. Each interference association representation vector is used as the basic data for constructing the sparse Bayesian compressed sensing correlation prior vector. The correlation prior vectors are arranged in the order of their corresponding edges to form a correlation prior vector set.

[0043] Based on the numerical magnitude of each sparse prior vector in the sparse prior vector set, construct the corresponding sparsity hyperparameters. Record the sparsity hyperparameters according to the order of the sparse prior vectors in the sparse prior vector set to form a sparsity hyperparameter set.

[0044] Based on the numerical correlation degree of each correlation prior vector in the correlation prior vector set, a corresponding correlation hyperparameter is constructed. The correlation hyperparameter is recorded in the order of the correlation prior vectors in the correlation prior vector set to form a correlation hyperparameter set.

[0045] Optionally, the construction of the frequency hopping consistency constraint includes:

[0046] Arrange the frequency hopping sequences recorded during the frequency hopping communication process in chronological order, and record the frequency hopping channel index corresponding to each frequency hopping cycle in sequence to form a set of frequency hopping sequences;

[0047] Based on the frequency hopping channel index of adjacent frequency hopping cycles in the frequency hopping sequence set, the frequency hopping channel index corresponding to the previous frequency hopping cycle and the frequency hopping channel index corresponding to the current frequency hopping cycle are established in a fixed manner to form a frequency hopping mapping relationship;

[0048] Using the frequency hopping mapping relationship as the cross-cycle mapping rule, the sparse coefficients of the previous frequency hopping cycle are mapped to the current frequency hopping cycle according to the frequency hopping mapping relationship, thus obtaining the cross-cycle sparse coefficient set.

[0049] The set of sparse coefficients across cycles is recorded in the order of the frequency hopping mapping relationship as the initial set of sparse coefficients for the current frequency hopping cycle;

[0050] The initial sparse coefficient set of the current frequency hopping cycle is paired one by one with the sparse coefficients of the previous frequency hopping cycle according to the frequency hopping mapping relationship to form a sparse coefficient pairing set.

[0051] The sparse coefficient pairing set is constructed as a frequency hopping consistency constraint in a fixed manner.

[0052] Optionally, the generation of the reconstructed Bluetooth signal and signal quality indicators includes:

[0053] The sparse prior vector set and the correlation prior vector set are input into the sparse Bayesian compressed sensing solution process in a fixed order. The sparsity hyperparameter set and the correlation hyperparameter set are used as input parameters to control the constraint strength of the sparse prior vector and the constraint strength of the correlation prior vector, respectively.

[0054] The frequency hopping consistency constraint is input into the sparse Bayesian compressed sensing solution process, and the frequency hopping mapping relationship is used as the cross-frequency band sparse coefficient mapping rule input into the solution process.

[0055] Based on the sparse prior vector set, the correlation prior vector set, the sparsity hyperparameter set, the correlation hyperparameter set, the frequency hopping consistency constraint and the frequency hopping mapping relationship, sparsity constraints, correlation constraints and continuity constraints are applied to the cross-frequency band sparse coefficients during the sparse Bayesian compressed sensing solution process, forming a constraint structure for performing the sparse Bayesian compressed sensing solution.

[0056] Based on the constraint structure, sparse Bayesian compressed sensing is performed to estimate the sparse coefficients of the Bluetooth signal affected by interference, and a set of sparse coefficient estimation results is obtained.

[0057] The sparse coefficient estimation results are input into the signal reconstruction process, and the reconstructed Bluetooth signal is generated according to the reconstruction rules of sparse Bayesian compressed sensing.

[0058] The reconstructed Bluetooth signal is input into the signal quality index calculation process to generate the reconstructed signal quality index.

[0059] Optionally, the generation of the frequency hopping sequence adjustment parameters, channel selection parameters, transmit power parameters, and coding parameters includes:

[0060] The reconstructed signal quality indicators are recorded in the order of frequency hopping cycles to form a sequence of reconstructed signal quality indicators.

[0061] The interference trend prediction results output by the graph neural network are recorded in the order of frequency hopping cycles to form a sequence of interference trend prediction results;

[0062] The reconstructed signal quality index sequence and the interference trend prediction result sequence are mapped according to the frequency hopping period to form a signal quality and interference trend correspondence sequence;

[0063] Based on the correspondence between the reconstructed signal quality index and the interference trend prediction results of each frequency hopping cycle in the signal quality and interference trend correspondence sequence, frequency hopping sequence adjustment data is constructed, and frequency hopping sequence adjustment parameters are calculated.

[0064] Based on the correspondence between the reconstructed signal quality index and the interference trend prediction results of each frequency hopping cycle in the signal quality and interference trend correspondence sequence, channel selection data is constructed and channel selection parameters are calculated.

[0065] Based on the correspondence between the reconstructed signal quality index and the interference trend prediction results of each frequency hopping cycle in the signal quality and interference trend correspondence sequence, transmit power data is constructed and transmit power parameters are calculated.

[0066] Based on the correspondence between the reconstructed signal quality index and the interference trend prediction results of each frequency hopping cycle in the signal quality and interference trend correspondence sequence, coding data is constructed and coding parameters are calculated.

[0067] Optionally, maintaining the stable signal transmission includes:

[0068] The frequency hopping sequence adjustment parameters are applied to the frequency hopping sequence in a fixed order;

[0069] The channel selection parameters are applied to the channel selection process in a fixed order;

[0070] The transmit power parameters are applied in a fixed order during the transmit power control process;

[0071] The encoding parameters are applied in a fixed order during the encoding configuration process;

[0072] The adjustment of frequency hopping sequence, channel selection, transmit power and coding method are combined in a fixed manner to form a wireless communication link adjustment structure;

[0073] Based on the wireless communication link adjustment structure, the wireless communication link of the Bluetooth headset is adjusted in real time. The frequency hopping sequence adjustment parameters, channel selection parameters, transmit power parameters and coding parameters are applied synchronously to the wireless communication link of the Bluetooth headset to maintain stable signal transmission of the Bluetooth headset under multi-source interference conditions.

[0074] The beneficial effects of this invention are:

[0075] This invention constructs multi-source radio frequency interference in Bluetooth frequency hopping communication environments as an interference topology with node-edge relationships. It then utilizes a graph neural network to perform neighborhood aggregation and deep representation learning on the node and edge features of this topology. This enables structured inference of the spatial distribution, coupling relationships, and trends of interference changes. Compared to traditional anti-interference methods that rely on single-point signal features or fixed threshold judgments, this invention can capture the correlation patterns between interferences from a global perspective. This allows the system to accurately identify interference sources and intensity changes under conditions of multi-device coexistence, high-density collisions, and complex environmental noise. Furthermore, this invention directly constructs the interference features output by the graph neural network into sparse priors and correlation priors in sparse Bayesian compressed sensing solutions. This allows the signal reconstruction process to utilize the structured patterns of interference, thereby improving the stability and accuracy of sparse coefficient estimation. Compared to existing Bluetooth anti-interference technologies based on traditional compressed sensing, this invention not only retains the advantages of sparse representation in damaged signal recovery but also overcomes the problems of insufficient priors and unstable models in traditional methods.

[0076] This invention analyzes the timing characteristics of frequency-hopping communication, introduces a frequency-hopping mapping relationship, and constructs frequency-hopping consistency constraints to ensure the continuity of sparse coefficients during frequency band switching. This solves the problems of discontinuous signal reconstruction and ineffective utilization of cross-band information in frequency-hopping scenarios. Based on the reconstructed high-quality Bluetooth signal, this invention further combines the quality indicators of the reconstructed signal with interference trend prediction results to generate four types of communication parameters: frequency-hopping sequence, channel selection, transmit power, and coding method. This enables real-time dynamic adjustment of the Bluetooth headset's wireless communication link. Through this multi-parameter collaborative adjustment mechanism, the Bluetooth headset can proactively optimize the communication path, power output, and coding method according to real-time interference conditions, ensuring stable and continuous signal transmission even under severe interference environments. This invention achieves structured, intelligent, and dynamic improvements in the three key aspects of interference identification, signal recovery, and link adjustment, forming a complete end-to-end RF interference suppression system. This significantly enhances the Bluetooth headset's anti-interference capability and communication reliability in complex environments. Attached Figure Description

[0077] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0078] Figure 1 This is an overall flowchart of a Bluetooth headset signal interference suppression system based on graph neural networks proposed in this invention;

[0079] Figure 2 This is a schematic diagram of a sparse Bayesian compressed sensing solution structure in a Bluetooth headset signal interference suppression system based on a graph neural network proposed in this invention.

[0080] Figure 3 This is a schematic diagram illustrating the dynamic adjustment process of the Bluetooth wireless communication link in a Bluetooth headset signal interference suppression system based on graph neural networks proposed in this invention. Detailed Implementation

[0081] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0082] refer to Figures 1-3 A Bluetooth headset signal interference suppression system based on graph neural networks includes:

[0083] The data acquisition and processing module is used to acquire wireless channel operation data of frequency hopping communication and perform normalization processing;

[0084] The interference topology construction module constructs an interference topology graph containing node attribute vectors and edge attribute vectors based on the running data.

[0085] The graph neural network inference module is used to generate interference representation vectors and interference-related representation vectors, and to perform trend inference to generate interference trend prediction results.

[0086] The prior generation module is used to construct sparse prior vectors and correlated prior vectors and generate corresponding hyperparameter sets.

[0087] The frequency hopping consistency construction module is used to establish frequency hopping mapping relationships and construct frequency hopping consistency constraints;

[0088] The signal reconstruction module performs sparse Bayesian compressed sensing to solve and reconstruct the Bluetooth signal based on prior, hyperparameter and frequency hopping consistency constraints;

[0089] The communication parameter generation module generates frequency hopping sequence adjustment parameters, channel selection parameters, transmit power parameters, and coding parameters based on the reconstructed signal quality indicators and interference trend prediction results.

[0090] The link adjustment module adjusts the Bluetooth headset's wireless communication link in real time and maintains stable transmission.

[0091] In this embodiment, the modules are interconnected using the following method:

[0092] The wireless channel operation data of the Bluetooth headset during frequency hopping communication is collected. The wireless channel operation data includes received signal strength, channel occupancy rate, signal-to-noise ratio, frequency band information of nearby devices, and multi-source interference energy distribution. The wireless channel operation data is normalized according to frequency index and time order.

[0093] An interference topology graph is constructed based on the normalized wireless channel operation data. Frequency hopping channels and identifiable interference sources are used as nodes, and the interference coupling relationship between nodes is used as edges. The node attribute vector of each node and the edge attribute vector of each edge are recorded so that the interference topology graph can fully express the spatial structure relationship of the interference.

[0094] The interference topology graph is input into the graph neural network, and neighborhood aggregation operation is performed on the node attribute vector and edge attribute vector to obtain the interference representation vector and interference association representation vector of the frequency hopping channel.

[0095] The interference representation vector is constructed as a sparse prior vector of sparse Bayesian compressed sensing, and the interference association representation vector is constructed as a correlation prior vector of sparse Bayesian compressed sensing. Based on the sparse prior vector and the correlation prior vector, a set of sparsity hyperparameters and a set of correlation hyperparameters are established.

[0096] A frequency hopping mapping relationship is established based on the frequency hopping sequence in the frequency hopping communication process. The sparse coefficients of the previous frequency hopping cycle are allocated to the current frequency hopping cycle according to the frequency hopping mapping relationship to obtain the sparse coefficients of the current frequency hopping cycle. A frequency hopping consistency constraint is constructed and applied during the solution process to ensure that the sparse coefficients remain continuous during the frequency band switching process.

[0097] Under the conditions of applying sparse prior vector, correlation prior vector and frequency hopping consistency constraints, and controlling the constraint strength of sparse prior vector with sparsity hyperparameter set and the constraint strength of correlation prior vector with correlation hyperparameter set, cross-frequency band sparse coefficient allocation is completed according to frequency hopping mapping relationship, sparse Bayesian compressed sensing solution is performed, sparse coefficient estimation and signal reconstruction are performed on the Bluetooth signal affected by interference, and the reconstructed Bluetooth signal and the reconstructed signal quality index are obtained.

[0098] Based on the reconstructed signal quality index and the interference trend prediction results output by the graph neural network, the frequency hopping sequence adjustment parameters, channel selection parameters, transmit power parameters and coding parameters are calculated.

[0099] The wireless communication link of the Bluetooth headset is adjusted in real time based on the frequency hopping sequence adjustment parameters, channel selection parameters, transmit power parameters, and coding parameters to maintain stable signal transmission of the Bluetooth headset under multi-source interference conditions.

[0100] In this embodiment, the construction of the interference topology map includes:

[0101] The normalized wireless channel operation data is arranged according to the frequency hopping channel index order, and each frequency hopping channel and each identifiable interference source is recorded as a node to form a node set.

[0102] Based on the normalized received signal strength data, channel occupancy data, signal-to-noise ratio data, neighboring device frequency band information data, and multi-source interference energy distribution data, a node attribute vector is constructed for each node in the node set, forming a node attribute vector set.

[0103] Interference coupling degree is generated based on the relationship between the received signal strength change, the channel occupancy rate change, and the interference energy change between any two nodes in the node attribute vector set. The interference coupling degree is then combined in a fixed order to construct an edge attribute vector, forming an edge attribute vector set.

[0104] Establish the connection relationship between nodes based on the node set and the edge attribute vector set, and establish an edge between each node and the corresponding node determined by the edge attribute vector to form an edge set;

[0105] A graph structure is constructed using a set of nodes, a set of edges, a set of node attribute vectors, and a set of edge attribute vectors. This graph structure is then labeled as an interference topology graph, which expresses the spatial structural relationship between the frequency hopping channel and the identifiable interference source.

[0106] In this embodiment, the generation of the interference characterization vector, the interference correlation characterization vector, and the interference trend prediction result includes:

[0107] The interference topology graph is input into the graph neural network. The set of nodes, the set of edges, the set of node attribute vectors, and the set of edge attribute vectors in the interference topology graph are input into the graph structure receiver of the graph neural network in a fixed order.

[0108] The node attribute vector of each node in the node attribute vector set is used as the initial node feature of that node, and the edge attribute vector of each edge in the edge attribute vector set is used as the initial edge feature of that edge, thus forming the initial node feature set and the initial edge feature set.

[0109] Based on the initial node feature set, a neighborhood aggregation operation is performed on each node at the neighborhood aggregation operation end of the graph neural network to form a set of node aggregation vectors;

[0110] The formation of the node aggregation vector set specifically includes: for each node in the interference topology graph, firstly, collecting the node attribute vectors of all neighboring nodes of the node, assigning weights to the node attribute vectors of each neighboring node according to the aforementioned interference coupling relationship, then performing a weighted summation of the node attribute vectors of each neighboring node according to the corresponding weights, combining the weighted summation vector with the node attribute vector of the current node in a fixed order to form the node aggregation vector of the node, and recording the node aggregation vectors of all nodes in node index order to form a node aggregation vector set;

[0111] The node aggregation vector corresponding to each node in the node aggregation vector set is fused with the attribute vectors of all edges connected to that node in the initial edge feature set in a fixed manner to form a node update vector set.

[0112] The formation of the node update vector set specifically includes: for each node in the interference topology graph, firstly collect all edge attribute vectors connected to the node, combine the node aggregation vector of the node with each edge attribute vector in a fixed order, and perform accumulation or sequential synthesis operations on the combined vector in the order of the edge attribute vectors to form the node update vector corresponding to the node. Repeat the above operation for all nodes, and arrange the node update vectors generated by each node in the order of the node set to form the node update vector set.

[0113] Replace the node attribute vector of each node in the node update vector set with the node update vector corresponding to that node to form the node attribute vector set after node update;

[0114] Record the node update vector corresponding to each node in the set of updated node attribute vectors as the interference characterization vector of the frequency hopping channel corresponding to that node, and form an interference characterization vector set.

[0115] Based on the updated set of node attribute vectors and the initial set of edge features, the edge attribute vector of each edge and the node update vectors of the two connected nodes are combined in a fixed order to form a set of interference association representation vectors.

[0116] Arrange the set of interference characterization vectors in order of frequency hopping cycle, record the interference characterization vectors corresponding to each frequency hopping cycle, and form an interference characterization sequence.

[0117] Based on the changing relationship of the interference characterization vector between each frequency hopping cycle in the interference characterization sequence, trend inference is performed, and the trend inference results are recorded in the order of frequency hopping cycles to form the interference trend prediction results.

[0118] In this embodiment, the establishment of the sparsity hyperparameter set and the correlation hyperparameter set includes:

[0119] The interference representation vectors are recorded in a fixed order, and each interference representation vector is used as the basic data for constructing sparse Bayesian compressed sensing sparse prior vectors. The sparse prior vectors are arranged in the order of their corresponding nodes to form a sparse prior vector set.

[0120] The interference association representation vectors are recorded in a fixed order. Each interference association representation vector is used as the basic data for constructing the sparse Bayesian compressed sensing correlation prior vector. The correlation prior vectors are arranged in the order of their corresponding edges to form a correlation prior vector set.

[0121] Based on the numerical magnitude of each sparse prior vector in the sparse prior vector set, construct the corresponding sparsity hyperparameters. Record the sparsity hyperparameters according to the order of the sparse prior vectors in the sparse prior vector set to form a sparsity hyperparameter set.

[0122] Based on the numerical correlation degree of each correlation prior vector in the correlation prior vector set, a corresponding correlation hyperparameter is constructed. The correlation hyperparameter is recorded in the order of the correlation prior vectors in the correlation prior vector set to form a correlation hyperparameter set.

[0123] In this embodiment, the construction of the frequency hopping consistency constraint includes:

[0124] Arrange the frequency hopping sequences recorded during the frequency hopping communication process in chronological order, and record the frequency hopping channel index corresponding to each frequency hopping cycle in sequence to form a set of frequency hopping sequences;

[0125] Based on the frequency hopping channel index of adjacent frequency hopping cycles in the frequency hopping sequence set, the frequency hopping channel index corresponding to the previous frequency hopping cycle and the frequency hopping channel index corresponding to the current frequency hopping cycle are established in a fixed manner to form a frequency hopping mapping relationship;

[0126] Using the frequency hopping mapping relationship as the cross-cycle mapping rule, the sparse coefficients of the previous frequency hopping cycle are mapped to the current frequency hopping cycle according to the frequency hopping mapping relationship, thus obtaining the cross-cycle sparse coefficient set.

[0127] The set of sparse coefficients across cycles is recorded in the order of the frequency hopping mapping relationship as the initial set of sparse coefficients for the current frequency hopping cycle;

[0128] The initial sparse coefficient set of the current frequency hopping cycle is paired one by one with the sparse coefficients of the previous frequency hopping cycle according to the frequency hopping mapping relationship to form a sparse coefficient pairing set.

[0129] The sparse coefficient pairing set is constructed as a frequency hopping consistency constraint in a fixed manner, so that the frequency hopping consistency constraint imposes a continuity requirement on the cross-period sparse coefficients during the solution process.

[0130] In this embodiment, the generation of the reconstructed Bluetooth signal and signal quality indicators includes:

[0131] The sparse prior vector set and the correlation prior vector set are input into the sparse Bayesian compressed sensing solution process in a fixed order. The sparsity hyperparameter set and the correlation hyperparameter set are used as input parameters to control the constraint strength of the sparse prior vector and the constraint strength of the correlation prior vector, respectively.

[0132] The frequency hopping consistency constraint is input into the sparse Bayesian compressed sensing solution process, and the frequency hopping mapping relationship is used as the cross-frequency band sparse coefficient mapping rule input into the solution process.

[0133] Based on the sparse prior vector set, the correlation prior vector set, the sparsity hyperparameter set, the correlation hyperparameter set, the frequency hopping consistency constraint and the frequency hopping mapping relationship, sparsity constraints, correlation constraints and continuity constraints are applied to the cross-frequency band sparse coefficients during the sparse Bayesian compressed sensing solution process, forming a constraint structure for performing the sparse Bayesian compressed sensing solution.

[0134] Based on the constraint structure, sparse Bayesian compressed sensing is performed to estimate the sparse coefficients of the Bluetooth signal affected by interference, and a set of sparse coefficient estimation results is obtained.

[0135] The generation of the sparse coefficient estimation result set specifically includes: performing sparse Bayesian compressed sensing solution; during the solution process, the Bluetooth signal affected by interference is mapped to observation data in a fixed manner; the sparse prior vector, correlation prior vector, sparsity hyperparameter set, correlation hyperparameter set and frequency hopping consistency constraint are combined into solution constraints in a fixed manner; and the sparse coefficients of the Bluetooth signal are estimated iteratively based on the solution constraints. In each iteration, the sparse coefficient estimate of the current iteration is calculated, and the sparse coefficient estimate is recorded in a fixed order to form the sparse coefficient estimation entry for the corresponding iteration. When the sparse Bayesian compressed sensing solution converges, the sparse coefficient estimation entries of all iterations are recorded in the iteration order to form the sparse coefficient estimation result set.

[0136] The sparse coefficient estimation results are input into the signal reconstruction process, and the reconstructed Bluetooth signal is generated according to the reconstruction rules of sparse Bayesian compressed sensing.

[0137] The generation of the reconstructed Bluetooth signal specifically includes: in the signal reconstruction process, firstly, the sparse coefficient estimation result set is input into the reconstruction process in the order of frequency hopping cycles, the sparse coefficient estimation value corresponding to each frequency hopping cycle is mapped to the frequency domain coefficients of the Bluetooth signal in a fixed manner, and the mapped frequency domain coefficients are combined with the reconstruction basis vectors in the reconstruction dictionary in a fixed manner according to the reconstruction dictionary required for reconstruction to form the reconstruction intermediate vector corresponding to each frequency hopping cycle. The reconstruction intermediate vectors of each frequency hopping cycle are subjected to inverse transformation operation in a fixed order to convert the reconstruction intermediate vectors in the frequency domain into Bluetooth signal segments in the time domain. The time domain Bluetooth signal segments obtained according to the frequency hopping cycle are spliced ​​together in the order of the frequency hopping sequence to form a continuous reconstructed Bluetooth signal.

[0138] The process of calculating the signal quality index of the reconstructed Bluetooth signal input signal involves combining the signal recovery error, energy distribution consistency, waveform continuity, and smoothness between frequency hopping bands of the Bluetooth signal in a fixed manner to obtain the reconstructed signal quality index.

[0139] In this embodiment, the generation of the frequency hopping sequence adjustment parameters, channel selection parameters, transmit power parameters, and coding parameters includes:

[0140] The reconstructed signal quality indicators are recorded in the order of frequency hopping cycles to form a sequence of reconstructed signal quality indicators.

[0141] The interference trend prediction results output by the graph neural network are recorded in the order of frequency hopping cycles to form a sequence of interference trend prediction results;

[0142] The reconstructed signal quality index sequence and the interference trend prediction result sequence are mapped according to the frequency hopping period to form a signal quality and interference trend correspondence sequence;

[0143] Based on the correspondence between the reconstructed signal quality index and the interference trend prediction result of each frequency hopping cycle in the signal quality and interference trend correspondence sequence, frequency hopping sequence adjustment data is constructed, and frequency hopping sequence adjustment parameters are calculated. The generation of frequency hopping sequence adjustment parameters includes: calculating the frequency hopping difference amount by matching the reconstructed signal quality index and the interference trend prediction result in a fixed manner based on the frequency hopping sequence adjustment data, and performing smoothing processing on the frequency hopping difference amount; determining the hopping order of the frequency hopping sequence based on the smoothed frequency hopping difference amount to form the frequency hopping sequence adjustment parameters.

[0144] Based on the correspondence between the reconstructed signal quality index and the interference trend prediction result of each frequency hopping cycle in the signal quality and interference trend correspondence sequence, channel selection data is constructed, and channel selection parameters are calculated. The generation of the channel selection parameters includes: based on the channel selection data, calculating the channel quality score by combining the reconstructed signal quality index and the interference trend prediction result in a fixed manner, determining the channel usage order based on the score result, and recording the channel usage order to form the channel selection parameters.

[0145] Based on the correspondence between the reconstructed signal quality index and the interference trend prediction result of each frequency hopping cycle in the signal quality and interference trend correspondence sequence, transmit power data is constructed and transmit power parameters are calculated. The generation of transmit power parameters includes: based on the transmit power data, calculating the interference intensity according to the reconstructed signal quality index and the interference trend prediction result in a fixed manner, calculating the transmit power adjustment amount based on the interference intensity, mapping the adjustment amount to the transmit power value, and forming transmit power parameters.

[0146] Based on the correspondence between the reconstructed signal quality index and the interference trend prediction result of each frequency hopping cycle in the signal quality and interference trend correspondence sequence, coding data is constructed and coding parameters are calculated. The generation of coding parameters includes: based on the coding data, performing coding stability evaluation on the reconstructed signal quality index and the interference trend prediction result in a fixed manner, and judging whether the coding method needs to be adjusted based on the evaluation result, recording the result of maintaining or adjusting the coding method, and forming coding parameters.

[0147] In this embodiment, maintaining stable signal transmission includes:

[0148] The frequency hopping sequence adjustment parameters are applied to the frequency hopping sequence in a fixed order, so that the frequency hopping sequence is adjusted according to the frequency hopping sequence adjustment parameters;

[0149] The channel selection parameters are applied to the channel selection process in a fixed order, so that the channel selection is adjusted according to the channel selection parameters.

[0150] The transmit power parameters are applied in a fixed order to the transmit power control process, so that the transmit power is adjusted according to the transmit power parameters;

[0151] The encoding parameters are applied in a fixed order during the encoding configuration process, so that the encoding method is adjusted according to the encoding parameters.

[0152] The adjustment of frequency hopping sequence, channel selection, transmit power and coding method are combined in a fixed manner to form a wireless communication link adjustment structure;

[0153] Based on the wireless communication link adjustment structure, the wireless communication link of the Bluetooth headset is adjusted in real time. The frequency hopping sequence adjustment parameters, channel selection parameters, transmit power parameters and coding parameters are applied synchronously to the wireless communication link of the Bluetooth headset to maintain stable signal transmission of the Bluetooth headset under multi-source interference conditions.

[0154] Example 1:

[0155] To verify the feasibility of this invention in practice, it was applied to a Bluetooth headset voice call scenario in a multi-story open-plan office building in a coastal city. This building houses a research and development center, open office areas, and multiple meeting rooms. Staff commonly use wireless LAN access devices, Bluetooth keyboards and mice, smartphones, and tablets, and simultaneously wear Bluetooth headsets from various brands for voice communication and remote conferencing. In actual use, issues frequently arise, including intermittent Bluetooth headset calls, static in the voice, fluctuating volume of the other party, and occasional disconnections between the headset and the phone or computer. These problems are particularly prevalent during morning rush hours and afternoon meetings, when the radio frequency environment is extremely congested. Traditional firmware relying on simple channel energy detection and fixed-threshold frequency hopping mechanisms struggles to avoid multi-source interference in a timely and accurate manner, resulting in a significant decline in the user's subjective call experience.

[0156] In this scenario, the Bluetooth headset signal interference suppression system of this invention is integrated into a Bluetooth headset and its accompanying terminal software for testing, enabling continuous operation during normal office use. The headset and terminal collaboratively collect wireless channel operation data during frequency hopping communication, including received signal strength, channel occupancy, signal-to-noise ratio, neighboring device frequency band information, and multi-source interference energy distribution. This data is then normalized within the system according to frequency index and time order. The normalized data is used to construct an interference topology graph, abstracting each frequency hopping channel and identifiable interference source as nodes, and abstracting the coupling relationships obtained from signal strength changes, channel occupancy, and interference energy changes as edges, thus forming a graph structure that reflects the spatial structure and correlation of interference. This interference topology graph is continuously input into a graph neural network. The graph neural network performs neighborhood aggregation and feature fusion on node attribute vectors and edge attribute vectors, generating interference representation vectors representing the interference state of each frequency hopping channel and interference correlation representation vectors characterizing the strength of the correlation between the channel and the interference source. This allows the system to infer the interference distribution and changing trends from a global perspective, rather than relying solely on channel quality at a single moment.

[0157] Based on interference topology inference, the system of this invention constructs the interference representation vector as a sparse prior vector in sparse Bayesian compressed sensing and the interference correlation representation vector as a correlation prior vector, thereby generating a sparsity hyperparameter set and a correlation hyperparameter set. When a Bluetooth headset makes a call in a high-interference environment, if there is local damage to the voice signal, incomplete buffer segments, or low signal-to-noise ratio in some frequency bands, the system will invoke the sparse Bayesian compressed sensing solution process, using the interfered Bluetooth voice signal as observation data, introducing the aforementioned sparse prior, correlation prior, and corresponding hyperparameters, and combining the frequency hopping mapping relationship and frequency hopping consistency constraints constructed based on the actual frequency hopping sequence, applying sparsity constraints, correlation constraints, and continuity constraints to the sparse coefficients across frequency bands, and iteratively estimating the sparse coefficients corresponding to each frequency hopping channel. With the help of this structured prior information, the system can recover the interference-damaged voice segments under frequency hopping communication, and convert the frequency domain sparse coefficients into a time domain voice signal through the reconstruction process, generating a continuous reconstructed Bluetooth signal.

[0158] Multiple meeting rooms and open office areas were selected in the aforementioned office building. Over a continuous workday, multiple rounds of voice calls and remote conferences were conducted. Some headsets used the existing firmware, while others used the system of this invention. Voice communication and conference calls were repeatedly conducted during the high-interference periods of the morning, noon, and evening. The stability of the Bluetooth connection, whether the call was interrupted, the user's subjective perception of voice clarity, and the continuity heard by the other party were recorded during the calls. These records were then compared and analyzed with the reconstructed signal quality indicators and interference trend prediction results generated internally by the system. It was observed that during periods with a significant increase in surrounding wireless devices and severe channel occupancy, the system of this invention was able to more promptly identify which frequency-hopping channels were affected by the superposition of multi-source interference. After signal reconstruction, it automatically generated frequency-hopping sequence adjustment parameters, channel selection parameters, transmit power parameters, and coding parameters based on signal quality indicators and interference trends. This dynamically adjusted the wireless communication link of the Bluetooth headsets, making them more inclined to choose channels with relatively less interference and more stable quality for frequency hopping. The transmit power was adjusted appropriately according to the strength of the interference, and the coding method was adaptively adjusted according to the current interference situation.

[0159] By analyzing and comparing operational records and data from multiple days and time periods in the aforementioned office scenarios, it was found that Bluetooth headsets using the system of this invention exhibit better call continuity and more coherent voice content in complex electromagnetic environments. Users experience smoother calls in multi-source interference environments, and voice interruptions and repetitions due to signal instability are significantly reduced in remote conferences. Combined with the trend of signal quality indicators generated internally by the system, it can be seen that the interference inference mechanism based on interference topology graphs and graph neural networks, as well as the signal reconstruction mechanism combining sparse Bayesian compressed sensing and frequency hopping consistency constraints proposed in this invention, have indeed played their expected role in real-world complex radio frequency environments. This has achieved end-to-end intelligent interference suppression of Bluetooth headset signals, verifying the feasibility and beneficial effects of this invention in engineering applications.

[0160] Table 1. Performance Comparison of the Invention System and Traditional Bluetooth Anti-interference Methods ; As shown in Table 1, in terms of call drop rate, this invention reduces the drop rate from 8.5% of the traditional method to 2.1%, a decrease of approximately 75%. This significant improvement is mainly due to the fact that this invention accurately identifies interference distribution through interference topology maps and graph neural networks, making frequency hopping sequences and channel selection more precise, thereby reducing call drops caused by sudden channel interference.

[0161] The subjective speech clarity score has been improved from the traditional 3.6 to 4.3, an improvement of about 19%. This invention not only reduces the intrusion of background interference on speech, but also improves the continuity and naturalness of speech waveform by sparsely reconstructing damaged speech segments, making the speech heard by the other end clearer and more coherent, and significantly improving the user's subjective experience.

[0162] The packet retransmission rate decreased from 12.8% to 4.7%, a reduction of approximately 63%. This is because the present invention can predict interference trends in advance and select a cleaner frequency hopping channel, resulting in more stable packet transmission and reduced duplicate transmissions.

[0163] Effective data throughput increased from 71% to 86%, an improvement of approximately 21%, indicating that this invention improves link utilization efficiency in complex interference environments. Dynamically adjusting transmit power and coding methods concentrates transmission resources on higher-quality channels, thereby increasing overall effective bandwidth.

[0164] The signal reconstruction success rate increased from 78% to 93%, an improvement of approximately 19%, indicating that the present invention can more effectively recover damaged speech data under multi-source interference. The interference representation and correlation prior provided by the graph neural network provide more accurate structured information for compressed sensing solutions, thereby improving the reconstruction quality.

[0165] As can be seen from the table data, the reason why this invention can comprehensively surpass traditional methods lies in integrating interference modeling, feature inference, sparse reconstruction, and frequency hopping consistency into a complete intelligent anti-interference link, enabling Bluetooth headsets to maintain clearer voice, lower interruption rate, and more stable transmission performance in complex radio frequency environments.

[0166] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A Bluetooth earphone signal interference suppression system based on a graph neural network, characterized in that, The method comprises the following steps: a data acquisition and processing module is used to acquire the wireless channel operation data of frequency hopping communication and perform normalization processing; an interference topology construction module is used to construct an interference topology graph containing node attribute vectors and edge attribute vectors according to the operation data; a graph neural network inference module is used to generate an interference feature vector and an interference association feature vector, and perform trend inference to generate an interference trend prediction result; a prior generation module is used to construct a sparse prior vector and a correlation prior vector and generate a corresponding hyperparameter set; a frequency hopping consistency construction module is used to establish a frequency hopping mapping relationship and construct a frequency hopping consistency constraint; a signal reconstruction module is used to perform sparse Bayesian compressed sensing solving according to the prior, hyperparameters and frequency hopping consistency constraints, and reconstruct the Bluetooth signal; a communication parameter generation module is used to generate frequency hopping sequence adjustment parameters, channel selection parameters, transmission power parameters and coding parameters according to the signal quality indicators and interference trend prediction results after reconstruction; a link adjustment module is used to perform real-time adjustment on the Bluetooth earphone wireless communication link and maintain stable transmission.

2. The Bluetooth earphone signal interference suppression system based on a graph neural network according to claim 1, wherein, The modules are realized through the following methods: wireless channel operation data of the Bluetooth earphone in the frequency hopping communication process is acquired, and the wireless channel operation data is normalized according to the frequency point index and time sequence; an interference topology graph is constructed according to the normalized wireless channel operation data; the interference topology graph is input into the graph neural network to obtain the interference feature vector and the interference association feature vector of the frequency hopping channel, and the trend inference is performed to generate the interference trend prediction result; the interference feature vector and the interference association feature vector are constructed into the sparse prior vector and the correlation prior vector of the sparse Bayesian compressed sensing, and the sparsity hyperparameter set and the correlation hyperparameter set are established; a frequency hopping mapping relationship is established according to the frequency hopping sequence to obtain the sparsity coefficient of the current frequency hopping period, and the frequency hopping consistency constraint is constructed; sparse Bayesian compressed sensing solving is performed to estimate the sparsity coefficient and reconstruct the signal of the Bluetooth signal affected by the interference, and the reconstructed Bluetooth signal and the signal quality indicators are obtained; the frequency hopping sequence adjustment parameters, the channel selection parameters, the transmission power parameters and the coding parameters are calculated according to the reconstructed signal quality indicators and the interference trend prediction results; the wireless communication link of the Bluetooth earphone is adjusted in real time to maintain stable signal transmission of the Bluetooth earphone under the condition of multi-source interference.

3. The Bluetooth earphone signal interference suppression system based on a graph neural network according to claim 2, characterized in that, The construction of the interference topology graph comprises: the normalized wireless channel operation data is arranged according to the frequency hopping channel index sequence, each frequency hopping channel and each identifiable interference source are recorded as nodes to form a node set; a node attribute vector is constructed for each node in the node set according to the normalized received signal strength data, channel occupancy rate data, signal-to-noise ratio data, adjacent device frequency band information data and multi-source interference energy distribution data to form a node attribute vector set; an interference coupling degree is generated according to the received signal strength change relationship, the channel occupancy rate change relationship and the interference energy change relationship between any two nodes in the node attribute vector set, and the interference coupling degree is combined to construct an edge attribute vector according to a fixed sequence to form an edge attribute vector set. According to the node set and the edge attribute vector set, a connection relationship between nodes is established, and an edge is established between each node and a corresponding node determined according to the edge attribute vector to form an edge set; The node set, the edge set, the node attribute vector set and the edge attribute vector set form a graph structure, and the graph structure is marked as an interference topology graph.

4. The Bluetooth earphone signal interference suppression system based on a graph neural network according to claim 2, characterized in that, The generation of the interference representation vector, the interference correlation representation vector and the interference trend prediction result includes: The interference topology graph is input into the graph neural network, and the node set, the edge set, the node attribute vector set and the edge attribute vector set in the interference topology graph are input into the graph structure receiving end of the graph neural network in a fixed order; Each node attribute vector in the node attribute vector set is taken as an initial node feature of the node, and each edge attribute vector in the edge attribute vector set is taken as an initial edge feature of the edge to form an initial node feature set and an initial edge feature set; According to the initial node feature set, neighborhood aggregation operation is performed on each node at the neighborhood aggregation operation end of the graph neural network to form a node aggregation vector set; The node aggregation vector corresponding to each node in the node aggregation vector set and all edge attribute vectors connected to the node in the initial edge feature set are fused in a fixed manner to form a node update vector set; The node update vector corresponding to each node in the node update vector set replaces the node attribute vector of the node to form a node attribute vector set after node update; The node update vector corresponding to each node in the node attribute vector set after node update is recorded as the interference representation vector of the corresponding frequency hopping channel of the node to form an interference representation vector set; According to the node attribute vector set after node update and the initial edge feature set, the edge attribute vector of each edge and the node update vectors of the two nodes connected thereto are combined in a fixed order to form an interference correlation representation vector set; The interference representation vector set is arranged in a frequency hopping period order, and the interference representation vectors corresponding to each frequency hopping period are recorded to form an interference representation sequence; According to the change relationship of the interference representation vectors between each frequency hopping period in the interference representation sequence, trend inference is performed, and the trend inference result is recorded in a frequency hopping period order to form an interference trend prediction result.

5. The Bluetooth earphone signal interference suppression system based on a graph neural network according to claim 2, characterized in that, The establishment of the sparsity hyperparameter set and the correlation hyperparameter set includes: Each interference representation vector is recorded in a fixed order, each interference representation vector is taken as basic data for constructing a sparse Bayesian compressed sensing sparse prior vector, and each sparse prior vector is arranged in a corresponding node order to form a sparse prior vector set; Each interference correlation representation vector is recorded in a fixed order, each interference correlation representation vector is taken as basic data for constructing a sparse Bayesian compressed sensing correlation prior vector, and each correlation prior vector is arranged in a corresponding edge order to form a correlation prior vector set; According to the numerical amplitude of each sparse prior vector in the sparse prior vector set, a corresponding sparsity hyperparameter is constructed, and the sparsity hyperparameters are recorded in the arrangement order of the sparse prior vectors in the sparse prior vector set to form a sparsity hyperparameter set; The corresponding correlation hyper-parameter is constructed according to the numerical correlation degree of each correlation prior vector in the correlation prior vector set, and the correlation hyper-parameter is recorded according to the arrangement order of the correlation prior vector in the correlation prior vector set to form a correlation hyper-parameter set.

6. The Bluetooth earphone signal interference suppression system based on a graph neural network according to claim 2, characterized in that, The construction of the frequency hopping consistency constraint includes: The frequency hopping sequence recorded in the frequency hopping communication process is arranged in time sequence, and the frequency hopping channel index corresponding to each frequency hopping period is recorded in sequence to form a frequency hopping sequence set; According to the frequency hopping channel index of adjacent frequency hopping periods in the frequency hopping sequence set, the corresponding relationship between the frequency hopping channel index corresponding to the previous frequency hopping period and the frequency hopping channel index corresponding to the current frequency hopping period is established in a fixed manner to form a frequency hopping mapping relationship; The frequency hopping mapping relationship is taken as a cross-period mapping rule, and the sparsity coefficient of the previous frequency hopping period is corresponded to the current frequency hopping period according to the frequency hopping mapping relationship to obtain a cross-period sparsity coefficient set; The cross-period sparsity coefficient set is recorded as the initial sparsity coefficient set of the current frequency hopping period according to the arrangement order of the frequency hopping mapping relationship; The initial sparsity coefficient set of the current frequency hopping period and the sparsity coefficient of the previous frequency hopping period are paired one by one according to the frequency hopping mapping relationship to form a sparsity coefficient pairing set; The sparsity coefficient pairing set is constructed into the frequency hopping consistency constraint in a fixed manner.

7. The Bluetooth earphone signal interference suppression system based on a graph neural network according to claim 2, characterized in that, The generation of the reconstructed Bluetooth signal and the signal quality index includes: The sparse prior vector set and the correlation prior vector set are input into the sparse Bayesian compressed sensing solving process in a fixed order, and the sparsity hyper-parameter set and the correlation hyper-parameter set are respectively taken as input parameters for controlling the constraint strength of the sparse prior vector and the correlation prior vector; The frequency hopping consistency constraint is input into the sparse Bayesian compressed sensing solving process, and the frequency hopping mapping relationship is taken as a cross-band sparsity coefficient mapping rule and input into the solving process; According to the sparse prior vector set, the correlation prior vector set, the sparsity hyper-parameter set, the correlation hyper-parameter set, the frequency hopping consistency constraint and the frequency hopping mapping relationship, the cross-band sparsity coefficient is subjected to sparsity constraint, correlation constraint and continuity constraint in the sparse Bayesian compressed sensing solving process to form a constraint structure for executing the sparse Bayesian compressed sensing solving; According to the constraint structure, the sparse Bayesian compressed sensing solving is executed, and the sparse coefficient estimation result set of the disturbed Bluetooth signal is obtained; The sparse coefficient estimation result set is input into the signal reconstruction process, and the reconstructed Bluetooth signal is generated according to the reconstruction rule of the sparse Bayesian compressed sensing; The reconstructed Bluetooth signal is input into the signal quality index calculation process to generate the reconstructed signal quality index.

8. The Bluetooth earphone signal interference suppression system based on a graph neural network according to claim 2, characterized in that, The generation of the frequency hopping sequence adjustment parameter, the channel selection parameter, the transmission power parameter and the coding parameter includes: The reconstructed signal quality index is recorded in sequence according to the frequency hopping period to form a reconstructed signal quality index sequence; The interference trend prediction result output by the graph neural network is recorded in sequence according to the frequency hopping period to form an interference trend prediction result sequence; The reconstructed signal quality index sequence and the interference trend prediction result sequence are corresponded according to the frequency hopping period to form a signal quality and interference trend corresponding sequence; According to the corresponding relationship between the reconstructed signal quality index and the interference trend prediction result of each frequency hopping period in the signal quality and interference trend corresponding sequence, frequency hopping sequence adjustment data is constructed, and frequency hopping sequence adjustment parameters are calculated; According to the corresponding relationship between the reconstructed signal quality index and the interference trend prediction result of each frequency hopping period in the signal quality and interference trend corresponding sequence, channel selection data is constructed, and channel selection parameters are calculated; According to the corresponding relationship between the reconstructed signal quality index and the interference trend prediction result of each frequency hopping period in the signal quality and interference trend corresponding sequence, transmission power data is constructed, and transmission power parameters are calculated; According to the corresponding relationship between the reconstructed signal quality index and the interference trend prediction result of each frequency hopping period in the signal quality and interference trend corresponding sequence, coding data is constructed, and coding parameters are calculated.

9. The Bluetooth earphone signal interference suppression system based on a graph neural network according to claim 2, wherein, The maintenance of the stable signal transmission includes: The frequency hopping sequence adjustment parameters are applied to the frequency hopping sequence in a fixed order; The channel selection parameters are applied to the channel selection process in a fixed order; The transmission power parameters are applied to the transmission power control process in a fixed order; The coding parameters are applied to the coding configuration process in a fixed order; The adjustment of the frequency hopping sequence, the adjustment of the channel selection, the adjustment of the transmission power, and the adjustment of the coding mode are combined in a fixed manner to form a wireless communication link adjustment structure; According to the wireless communication link adjustment structure, real-time adjustment is performed on the wireless communication link of the Bluetooth headset, and the frequency hopping sequence adjustment parameters, the channel selection parameters, the transmission power parameters, and the coding parameters act on the wireless communication link of the Bluetooth headset synchronously, thereby maintaining the stable signal transmission of the Bluetooth headset under the condition of multi-source interference.

Images