A magnetic communication enhancement method and apparatus based on motion posture and interference recognition

Abstract

This invention provides a magnetic communication enhancement method and apparatus based on motion posture and interference recognition, relating to the field of positioning technology. The method includes: constructing a magnetic communication time-series dataset and inputting it into a bidirectional data transfer module to obtain inputs for an inverse signal query network and a timing enhancement network; obtaining the output features of the inverse signal query network based on the inputs and the RepViT-based inverse signal query network; and obtaining antenna gain features based on the inputs, the output features of the inverse signal query network, and the timing enhancement network. This invention trains the magnetic communication enhancement network using communication signals from a communication terminal with a transmit array, receive array, and IMU under different postures and paths, obtaining the optimal gain of the receiving and transmit arrays of the communication terminal under the current posture. This significantly reduces the impact of target motion posture and surrounding interference signals, and adaptively adjusts the gain of different transmit and receive antennas based on the learning results, improving communication quality.

Description

Technical Field

[0001] This invention relates to the field of positioning technology, and in particular to a magnetic communication enhancement method and apparatus based on motion posture and interference recognition. Background Technology

[0002] Magnetic field communication is a communication method that uses magnetic fields as the information transmission carrier. It is currently widely used in smart IoT, industrial automation, biomedicine, security monitoring, and energy management. Because magnetic field communication generally uses relatively low communication frequencies, it may be subject to magnetic field interference from electrical equipment, power lines, and the Earth's magnetic field. Methods such as shielding, proper circuit layout design, differential-mode signal transmission, and digital signal processing are typically employed to reduce the impact of interference signals. Furthermore, magnetic field communication is directional; the relative position and orientation of the transmitter and receiver affect the communication performance. Therefore, appropriate measures and technical means are needed to address issues such as magnetic field interference and directionality, thereby improving the reliability and stability of magnetic field communication. Summary of the Invention

[0003] To address the technical problems of magnetic field interference and directionality in existing technologies, this invention provides a magnetic communication enhancement method and apparatus based on motion posture and interference identification. The technical solution is as follows:

[0004] On the one hand, a magnetic communication enhancement method based on motion posture and interference recognition is provided. This method is implemented by a magnetic communication enhancement device and includes:

[0005] S1. Construct a magnetic communication time series dataset; the magnetic communication time series dataset includes IMU output, receiver array channel quality parameters and optimal array gain under different poses.

[0006] S2. Input the magnetic communication time series dataset into the bidirectional data conversion module to obtain the input of the inverse signal query network and the input of the time series enhancement network.

[0007] S3. Design an inverse signal query network based on the RepViT network. Obtain the output characteristics of the inverse signal query network based on the input and output characteristics of the inverse signal query network.

[0008] S4. Based on the input of the timing enhancement network, the output characteristics of the inverse signal query network, and the timing enhancement network, the antenna gain characteristics are obtained.

[0009] S5. Construct a magnetic communication enhancement network based on the bidirectional digital conversion module, the inverse signal query network, and the timing enhancement network. Obtain the output of the magnetic communication enhancement network based on the output characteristics of the inverse signal query network, the antenna gain characteristics, and the magnetic communication enhancement network.

[0010] S6. Train the magnetic communication enhancement network based on the magnetic communication time series dataset to obtain the trained magnetic communication enhancement model.

[0011] S7. Obtain the pose data of the magnetic communication enhancement to be performed, input the pose data into the trained magnetic communication enhancement model, and obtain the optimal array gain of the receiving array and transmitting array of the communication terminal under the current pose.

[0012] Optionally, the construction of the magnetic communication time series dataset in S1 includes:

[0013] The system acquires the IMU output, receiver array channel quality parameters, and optimal array gain of the receiver and transmitter arrays of the communication terminal when the terminal moves in different poses and along different acquisition paths.

[0014] The communication terminal carries a transmitter array, a receiver array, and an IMU.

[0015] Different types of interference signals are introduced into different sections of the acquisition path.

[0016] The data dimension of the IMU output is 1*9; the data dimension of the receive array channel quality parameters is N*M, where N represents the number of sensors and M is the number of channel parameters for each sensor; the data dimension of the optimal array gain is N+K, where K is the number of transmit antennas.

[0017] Optionally, in S2, the magnetic communication time-series dataset is input to the bidirectional data conversion module to obtain the inputs to the inverse signal query network and the time-series enhancement network, including:

[0018] The channel quality parameters of the receiving array are expanded to obtain a 1*(N*M) dimensional vector. This expanded vector is then concatenated with the IMU output to obtain the input C of the 1*(N*M+9) dimensional timing enhancement network. mt Where N represents the number of sensors, and M is the number of channel parameters for each sensor.

[0019] The elements in the IMU output are multiplied element by element with the channel quality parameters of the receiving array to obtain 9 matrix channels. The 9 matrix channels are then concatenated to obtain the input C of the N*M*9 dimensional inverse signal query network.

[0020] Optionally, in S3, an inverse signal query network based on the RepViT network is designed. Based on the input and output characteristics of the inverse signal query network, the following features are obtained:

[0021] The input C of the inverse signal query network is fed into a convolutional layer with a 1*1 kernel to obtain an N*M*3 dimensional output matrix; where N represents the number of sensors and M is the number of channel parameters for each sensor.

[0022] The N*M*3 dimensional output matrix is ​​input into the RepViT network for magnetic signal feature extraction, resulting in the output feature CO of the inverse signal query network with dimension 1*(N+K); where K is the number of transmit antennas.

[0023] Optionally, the temporal augmentation network includes a forget gate, an input gate, and a gain gate.

[0024] S4 obtains antenna gain characteristics based on the input of the timing enhancement network, the output characteristics of the inverse signal query network, and the timing enhancement network itself, including:

[0025] S41. Obtain the optimal antenna gain G from the previous time t-1. t-1 And the input C of the time-series augmentation network at the current time t. mt The concatenation is performed in the transpose layer of the forget gate to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose matrix to obtain the feature matrix G. M1 For the characteristic matrix G M1 Perform a convolution operation to obtain the convolutional feature matrix G. M2 The sigmoid activation function is used on the feature matrix G. M2 Activation is performed to obtain the feature matrix G after sigmoid activation. M2 The feature matrix G activated by sigmoid M2 With gain memory cells S t-1 Multiplying these values ​​yields the output gain of the forgetting gate, S, for the memory cell. t1 .

[0026] S42, Obtain the optimal antenna gain G from the previous time step t-1. t-1 And the input C of the time-series augmentation network at the current time t. mt In the first branch of the input gate, concatenation is performed to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the feature matrix G. M3 For the characteristic matrix G M3 Perform a convolution operation to obtain the convolutional feature matrix G. M4 The sigmoid activation function is used on the feature matrix G. M4 Activation is performed to obtain the feature matrix G after sigmoid activation. M4 .

[0027] The optimal antenna gain G obtained at the previous time t-1 t-1 And the input C of the time-series augmentation network at the current time t. mtIn the second branch of the input gate, concatenation is performed to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the feature matrix G. M5 For the characteristic matrix G M5 Perform a convolution operation to obtain the convolutional feature matrix G. M6 The Tanh activation function is used on the feature matrix G. M6 Activation is performed to obtain the feature matrix G after Tanh activation. M6 .

[0028] The feature matrix G after sigmoid activation M4 With the feature matrix G activated by Tanh M6 Multiply the result and add it to the output gain of the forget gate, memory cell S. t1 Element-wise summation is performed, and the summation result is passed through a fully connected layer to obtain the output gain of the input gate, memory cell S. t .

[0029] S43, Obtain the optimal antenna gain G from the previous time step t-1. t-1 And the input C of the time-series augmentation network at the current time t. mt The input is fed into the first branch of the gain gate for concatenation, resulting in a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the eigenvector matrix G. M7 For the characteristic matrix G M7 Perform a convolution operation to obtain the convolutional feature matrix G. M8 The sigmoid activation function is used on the feature matrix G. M8 Activation is performed to obtain the feature matrix G after sigmoid activation. M8 .

[0030] The output gain of the input gate is obtained by the memory cell S. t The inverse signal query network output feature CO is input into the second branch of the gain gate for concatenation to obtain a concatenated one-dimensional vector. The concatenated one-dimensional vector is multiplied by the transpose of the concatenated one-dimensional vector to obtain the multiplied feature matrix CO1. The feature matrix CO1 is convolved to obtain the convolved feature matrix CO2. The Tanh activation function is used to activate the feature matrix CO2 to obtain the Tanh-activated feature matrix CO2.

[0031] The feature matrix G after sigmoid activation M8 Multiplying the feature matrix CO2 after Tanh activation and passing the result through a fully connected layer, we obtain the optimal antenna gain feature G at time t. t .

[0032] Optionally, in S5, the magnetic communication enhancement network output is obtained based on the inverse signal query network output characteristics, antenna gain characteristics, and the magnetic communication enhancement network, including:

[0033] The inverse signal query network output feature CO and antenna gain feature G are used to determine the inverse signal query network output feature CO and antenna gain feature G. t Multiplying these matrices yields a (N+K)*(N+K) dimensional feature matrix. This feature matrix is ​​then processed through a max-pooling layer and a fully connected layer to obtain the output of a magnetic communication enhancement network with a dimension of 1*(N+K).

[0034] In the output of the magnetic communication enhancement network, each element represents the optimal array gain CG of the receiving array and the transmitting array of the communication terminal under the current pose.

[0035] Optionally, in S6, the magnetic communication enhancement network is trained based on the magnetic communication time-series dataset to obtain a trained magnetic communication enhancement model, including:

[0036] The inverse signal query network is trained using the magnetic communication time series dataset to obtain the trained inverse signal query network.

[0037] The internal parameters of the pre-trained inverse signal query network are fixed, and the time-series augmentation network is trained based on the magnetic communication time-series dataset to obtain the pre-trained time-series augmentation network.

[0038] The trained inverse signal query network and the trained time series enhancement network are used as pre-trained models. The magnetic communication enhancement network is trained based on the magnetic communication time series dataset to obtain the trained magnetic communication enhancement model.

[0039] On the other hand, a magnetic communication enhancement device based on motion posture and interference recognition is provided. This device is applied to a magnetic communication enhancement method based on motion posture and interference recognition, and includes:

[0040] The magnetic communication timing dataset construction module is used to construct a magnetic communication timing dataset; the magnetic communication timing dataset includes IMU output, receiver array channel quality parameters and optimal array gain under different poses.

[0041] The bidirectional data transfer module is used to input the magnetic communication time series dataset into the bidirectional data transfer module to obtain the input of the inverse signal query network and the input of the time series enhancement network.

[0042] The inverse signal query network module is used to design an inverse signal query network based on the RepViT network. Based on the input and output characteristics of the inverse signal query network, the module obtains the output characteristics of the inverse signal query network.

[0043] The timing enhancement network module is used to obtain antenna gain characteristics based on the input of the timing enhancement network, the output characteristics of the inverse signal query network, and the timing enhancement network itself.

[0044] The magnetic communication enhancement network module is used to construct a magnetic communication enhancement network based on the bidirectional digital conversion module, the inverse signal query network, and the timing enhancement network. The output of the magnetic communication enhancement network is obtained based on the output characteristics of the inverse signal query network, the antenna gain characteristics, and the magnetic communication enhancement network.

[0045] The training module is used to train the magnetic communication enhancement network based on the magnetic communication time series dataset to obtain the trained magnetic communication enhancement model.

[0046] The output module is used to acquire the pose data of the magnetic communication enhancement to be performed, input the pose data into the trained magnetic communication enhancement model, and obtain the optimal array gain of the receiving array and transmitting array of the communication terminal under the current pose.

[0047] Optionally, the magnetic communication time series dataset construction module is further used for:

[0048] The system acquires the IMU output, receiver array channel quality parameters, and optimal array gain of the receiver and transmitter arrays of the communication terminal when the terminal moves in different poses and along different acquisition paths.

[0049] The communication terminal carries a transmitter array, a receiver array, and an IMU.

[0050] Different types of interference signals are introduced into different sections of the acquisition path.

[0051] The data dimension of the IMU output is 1*9; the data dimension of the receive array channel quality parameters is N*M, where N represents the number of sensors and M is the number of channel parameters for each sensor; the data dimension of the optimal array gain is N+K, where K is the number of transmit antennas.

[0052] Optionally, the bidirectional digital-to-digital converter module is further used for:

[0053] The channel quality parameters of the receiving array are expanded to obtain a 1*(N*M) dimensional vector. This expanded vector is then concatenated with the IMU output to obtain the input C of the 1*(N*M+9) dimensional timing enhancement network. mt Where N represents the number of sensors, and M is the number of channel parameters for each sensor.

[0054] The elements in the IMU output are multiplied element by element with the channel quality parameters of the receiving array to obtain 9 matrix channels. The 9 matrix channels are then concatenated to obtain the input C of the N*M*9 dimensional inverse signal query network.

[0055] Optionally, the inverse signal query network module is further used for:

[0056] The input C of the inverse signal query network is fed into a convolutional layer with a 1*1 kernel to obtain an N*M*3 dimensional output matrix; where N represents the number of sensors and M is the number of channel parameters for each sensor.

[0057] The N*M*3 dimensional output matrix is ​​input into the RepViT network for magnetic signal feature extraction, resulting in the output feature CO of the inverse signal query network with dimension 1*(N+K); where K is the number of transmit antennas.

[0058] Optionally, the temporal augmentation network includes a forget gate, an input gate, and a gain gate.

[0059] The timing enhancement network module is further used for:

[0060] S41. Obtain the optimal antenna gain G from the previous time t-1. t-1 And the input C of the time-series augmentation network at the current time t. mt The concatenation is performed in the transpose layer of the forget gate to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose matrix to obtain the feature matrix G. M1 For the characteristic matrix G M1 Perform a convolution operation to obtain the convolutional feature matrix G. M2 The sigmoid activation function is used on the feature matrix G. M2 Activation is performed to obtain the feature matrix G after sigmoid activation. M2 The feature matrix G activated by sigmoid M2 With gain memory cells S t-1 Multiplying these values ​​yields the output gain of the forgetting gate, S, for the memory cell. t1 .

[0061] S42, Obtain the optimal antenna gain G from the previous time step t-1. t-1 And the input C of the time-series augmentation network at the current time t. mt In the first branch of the input gate, concatenation is performed to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the feature matrix G. M3 For the characteristic matrix G M3 Perform a convolution operation to obtain the convolutional feature matrix G. M4 The sigmoid activation function is used on the feature matrix G. M4 Activation is performed to obtain the feature matrix G after sigmoid activation. M4 .

[0062] The optimal antenna gain G obtained at the previous time t-1 t-1 And the input C of the time-series augmentation network at the current time t. mt In the second branch of the input gate, concatenation is performed to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the feature matrix G. M5 For the characteristic matrix G M5 Perform a convolution operation to obtain the convolutional feature matrix G. M6 The Tanh activation function is used on the feature matrix G. M6 Activation is performed to obtain the feature matrix G after Tanh activation. M6 .

[0063] The feature matrix G after sigmoid activation M4 With the feature matrix G activated by Tanh M6 Multiply the result and add it to the output gain of the forget gate, memory cell S. t1 Element-wise summation is performed, and the summation result is passed through a fully connected layer to obtain the output gain of the input gate, memory cell S. t .

[0064] S43, Obtain the optimal antenna gain G from the previous time step t-1. t-1 And the input C of the time-series augmentation network at the current time t. mt The input is fed into the first branch of the gain gate for concatenation, resulting in a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the eigenvector matrix G. M7 For the characteristic matrix G M7 Perform a convolution operation to obtain the convolutional feature matrix G. M8 The sigmoid activation function is used on the feature matrix G. M8 Activation is performed to obtain the feature matrix G after sigmoid activation. M8 .

[0065] The output gain of the input gate is obtained by the memory cell S. t The inverse signal query network output feature CO is input into the second branch of the gain gate for concatenation to obtain a concatenated one-dimensional vector. The concatenated one-dimensional vector is multiplied by the transpose of the concatenated one-dimensional vector to obtain the multiplied feature matrix CO1. The feature matrix CO1 is convolved to obtain the convolved feature matrix CO2. The Tanh activation function is used to activate the feature matrix CO2 to obtain the Tanh-activated feature matrix CO2.

[0066] The feature matrix G after sigmoid activation M8Multiplying the feature matrix CO2 after Tanh activation and passing the result through a fully connected layer, we obtain the optimal antenna gain feature G at time t. t .

[0067] Optionally, the magnetic communication enhancement network module is further used for:

[0068] The inverse signal query network output feature CO and antenna gain feature G are used to determine the inverse signal query network output feature CO and antenna gain feature G. t Multiplying these matrices yields a (N+K)*(N+K) dimensional feature matrix. This feature matrix is ​​then processed through a max-pooling layer and a fully connected layer to obtain the output of a magnetic communication enhancement network with a dimension of 1*(N+K).

[0069] In the output of the magnetic communication enhancement network, each element represents the optimal array gain CG of the receiving array and the transmitting array of the communication terminal under the current pose.

[0070] Optionally, the training module is further used for:

[0071] The inverse signal query network is trained using the magnetic communication time series dataset to obtain the trained inverse signal query network.

[0072] The internal parameters of the pre-trained inverse signal query network are fixed, and the time-series augmentation network is trained based on the magnetic communication time-series dataset to obtain the pre-trained time-series augmentation network.

[0073] The trained inverse signal query network and the trained time series enhancement network are used as pre-trained models. The magnetic communication enhancement network is trained based on the magnetic communication time series dataset to obtain the trained magnetic communication enhancement model.

[0074] On the other hand, a magnetic communication enhancement device is provided, comprising: a processor; a memory storing computer-readable instructions, which, when executed by the processor, implement any of the methods described above for magnetic communication enhancement based on motion posture and interference recognition.

[0075] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement any of the above-described methods for enhancing magnetic communication based on motion posture and interference recognition.

[0076] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0077] In this invention, the magnetic communication enhancement method based on motion posture and interference recognition has a core technology of magnetic communication enhancement network. Compared with other traditional methods, the calculation process is real-time and accurate, which can significantly reduce the influence of target motion posture and surrounding interference signals. Furthermore, it can adaptively adjust the gain of different transceiver antennas based on the learning results to improve communication quality.

[0078] This magnetic communication enhancement network is a dual-branch composite neural network. The inverse signal query network has the powerful feature fitting ability of convolutional structures, and the temporal enhancement network is suitable for capturing long-term dependencies in temporal communication signals. Furthermore, the temporal enhancement network innovatively introduces convolutional features to guide the convergence of temporal features, realizing information interaction and deep fusion of dual-branch networks. Compared with single network structures or independent dual-branch structures, this structure can more comprehensively and accurately explore the magnetic sensing mechanism, thereby improving the accuracy of optimal antenna gain evaluation. Attached Figure Description

[0079] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0080] Figure 1 This is a flowchart of a magnetic communication enhancement method based on motion posture and interference identification provided in an embodiment of the present invention;

[0081] Figure 2 This is a network structure diagram of the forget gate provided in an embodiment of the present invention;

[0082] Figure 3 This is a network structure diagram of the input gate provided in an embodiment of the present invention;

[0083] Figure 4 This is a network structure diagram of the gain gate provided in an embodiment of the present invention;

[0084] Figure 5 This is a network structure diagram of the magnetic communication enhancement model provided in the embodiments of the present invention;

[0085] Figure 6 This is a flowchart illustrating the construction process of the magnetic communication enhancement model provided in this embodiment of the invention;

[0086] Figure 7 This is a block diagram of a magnetic communication enhancement device based on motion posture and interference recognition provided in an embodiment of the present invention;

[0087] Figure 8 This is a schematic diagram of the structure of a magnetic communication enhancement device provided in an embodiment of the present invention. Detailed Implementation

[0088] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0089] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0090] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0091] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0092] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0093] This invention provides a magnetic communication enhancement method based on motion posture and interference recognition. This method can be implemented by a magnetic communication enhancement device, which can be a terminal or a server. Figure 1 The flowchart shown is for a magnetic communication enhancement method based on motion posture and interference recognition. The processing flow of this method may include the following steps:

[0094] S1. Construct a magnetic communication time series dataset; the magnetic communication time series dataset includes IMU output, receiver array channel quality parameters and optimal array gain under different poses.

[0095] Optionally, step S1 above may include:

[0096] The system acquires the IMU output, receiver array channel quality parameters, and optimal array gain of the receiver and transmitter arrays of the communication terminal when the terminal moves in different poses and along different acquisition paths.

[0097] The communication terminal carries a transmitter array, a receiver array, and an IMU.

[0098] Different types of interference signals are introduced into different sections of the acquisition path.

[0099] The data dimension of the IMU output is 1*9; the data dimension of the receive array channel quality parameters is N*M, where N represents the number of sensors and M is the number of channel parameters for each sensor; the data dimension of the optimal array gain is N+K, where K is the number of transmit antennas.

[0100] In one feasible implementation, a magnetic communication data acquisition experiment is conducted. A communication terminal carrying a transmitting array, receiving array, and IMU is guided to move in different postures and along different acquisition paths, while various types of interference signals are introduced at different movement points. Real-time IMU (Inertial Measurement Unit) outputs (data dimension 1×9, including velocity, acceleration, and magnetic field strength in each direction) and channel quality parameters of the receiving array (data dimension N×M, where N represents the number of sensors and M is the number of channel parameters for each sensor, including signal strength, signal-to-noise ratio, bit error rate, etc.) are recorded simultaneously. The optimal gain of the receiving and transmitting arrays of the communication terminal under the current posture is also adjusted and recorded (data dimension N+K, where K is the number of transmitting antennas). After obtaining a sufficient amount of communication data, the IMU outputs, receiving array channel quality parameters, and optimal array gains under each posture are mapped one-to-one to form a magnetic communication time-series dataset.

[0101] S2. Input the magnetic communication time series dataset into the bidirectional data conversion module to obtain the input of the inverse signal query network and the input of the time series enhancement network.

[0102] Optionally, step S2 above may include:

[0103] The channel quality parameters of the receiving array are expanded to obtain a 1*(N*M) dimensional vector. This expanded vector is then concatenated with the IMU output to obtain the input C of the 1*(N*M+9) dimensional timing enhancement network. mt Where N represents the number of sensors, and M is the number of channel parameters for each sensor.

[0104] The elements in the IMU output are multiplied element by element with the channel quality parameters of the receiving array to obtain 9 matrix channels. The 9 matrix channels are then concatenated to obtain the input C of the N*M*9 dimensional inverse signal query network.

[0105] In one feasible implementation, the bidirectional data transfer module (BJT) processes the sensor data obtained from the communication terminal in real time into the input format for the inverse signal lookup network (ISGN) and the timing enhancement network. In the BJT, the N×M data-dimensional receiver array output signal is expanded into a 1*(N*M) dimensional vector and concatenated with the 1*9 dimensional IMU output to obtain a 1*(N*M+9) dimensional vector Cmt, which serves as the input to the timing enhancement network. The nine elements of the IMU output are multiplied element-wise with the receiver array signal matrix, and the resulting nine matrix channels are concatenated to obtain an N*M*9 dimensional matrix C, which serves as the input to the ISGN. This constitutes the internal operation of the BJT.

[0106] S3. Design an inverse signal query network based on the RepViT network. Obtain the output characteristics of the inverse signal query network based on the input and output characteristics of the inverse signal query network.

[0107] Optionally, step S3 above may include:

[0108] The input C of the inverse signal query network is fed into a convolutional layer with a 1*1 kernel to obtain an N*M*3 dimensional output matrix; where N represents the number of sensors and M is the number of channel parameters for each sensor.

[0109] The N*M*3 dimensional output matrix is ​​input into the RepViT network for magnetic signal feature extraction, resulting in the output feature CO of the inverse signal query network with dimension 1*(N+K); where K is the number of transmit antennas.

[0110] In one feasible implementation, since the magnetic communication enhancement method needs to be deployed to a mobile terminal for data processing, which places high demands on data latency, a novel lightweight backbone network, RepViT, is used as the inverse signal query main network. First, an N*M*9 dimensional data matrix C is input into a convolutional layer with a 1*1 kernel, transforming it into an N*M*3 dimensional matrix. Then, this matrix is ​​input into the RepViT network for magnetic signal feature extraction, where the output C4 of the RepViT network structure is set to (N+K), resulting in a one-dimensional vector CO with dimensions 1*(N+K). The inverse signal query network relies on the strong fitting ability of layer-by-layer convolution to extract the effective information from the current sensing data into the one-dimensional vector CO.

[0111] S4. Based on the input of the timing enhancement network, the output characteristics of the inverse signal query network, and the timing enhancement network, the antenna gain characteristics are obtained.

[0112] Optionally, the design of a time-enhanced network mainly includes three parts: forget gate design, input gate design, and gain gate design.

[0113] Step S4 above may include the following steps S41-S43:

[0114] S41. Obtain the optimal antenna gain G from the previous time t-1. t-1 And the input C of the time-series augmentation network at the current time t. mt The concatenation is performed in the transpose layer of the forget gate to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose matrix to obtain the feature matrix G. M1 For the characteristic matrix G M1 Perform a convolution operation to obtain the convolutional feature matrix G. M2 The sigmoid activation function is used on the feature matrix G. M2 Activation is performed to obtain the feature matrix G after sigmoid activation. M2 The feature matrix G activated by sigmoid M2 With gain memory cells S t-1 Multiplying these values ​​yields the output gain of the forgetting gate, S, for the memory cell. t1 .

[0115] One feasible implementation method is, for example Figure 2 As shown, the input to the forget gate is the optimal antenna gain G from the previous time step. t-1 Its dimension is 1*(N+K), and the sensing data C output by the bidirectional data conversion module at the current moment is... mt Its dimension is 1*(N*M+9). In the transpose layer, the two one-dimensional inputs are concatenated to obtain a 1*(N+K+N*M+9) dimensional one-dimensional vector; then this vector is multiplied by its own transpose matrix to obtain the (N+K+N*M+9)*(N+K+N*M+9) dimensional feature matrix G. M1 For G M1 Perform convolution operations to extract features and adjust dimensions, resulting in an (N+K)*(N+K) dimensional feature matrix G. M2 Subsequently, sigmoid was used to apply G M2 Activate. G at this moment. M2 The interference and redundancy information in the input signal are abstracted through convolution, and the corresponding positions of the redundancy information are set to 0. Then, the G signal activated by sigmoid is... M2 With gain memory cells S t-1 Multiplication is performed, thereby eliminating the gain memory cell S. t-1 From information irrelevant to the current gain adjustment task, we obtain (N+K)*(N+K) dimensional features S. t1 .

[0116] S42, Obtain the optimal antenna gain G from the previous time step t-1. t-1 And the input C of the time-series augmentation network at the current time t. mtIn the first branch of the input gate, concatenation is performed to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the feature matrix G. M3 For the characteristic matrix G M3 Perform a convolution operation to obtain the convolutional feature matrix G. M4 The sigmoid activation function is used on the feature matrix G. M4 Activation is performed to obtain the feature matrix G after sigmoid activation. M4 .

[0117] The optimal antenna gain G obtained at the previous time t-1 t-1 And the input C of the time-series augmentation network at the current time t. mt In the second branch of the input gate, concatenation is performed to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the feature matrix G. M5 For the characteristic matrix G M5 Perform a convolution operation to obtain the convolutional feature matrix G. M6 The Tanh activation function is used on the feature matrix G. M6 Activation is performed to obtain the feature matrix G after Tanh activation. M6 .

[0118] The feature matrix G after sigmoid activation M4 With the feature matrix G activated by Tanh M6 Multiply the result and add it to the output gain of the forget gate, memory cell S. t1 Element-wise summation is performed, and the summation result is passed through a fully connected layer to obtain the output gain of the input gate, memory cell S. t .

[0119] In one feasible implementation, the input to the input gate is the same as that of the forget gate, which is the optimal antenna gain G from the previous time step. t-1 Its dimension is 1*(N+K), and the sensing data C output by the bidirectional data conversion module at the current moment is... mt Its dimension is 1*(N*M+9). The network structure of the input gate is as follows: Figure 3 As shown, in the left branch, first adjust the antenna gain G. t-1 With sensor data G t-1 By inputting a transpose layer, a convolutional layer, and a sigmoid activation layer with a structure similar to that of a forget gate, a matrix G of dimension (N+K+N*M+9)*(N+K+N*M+9) is obtained sequentially. M3 The (N+K)*(N+K) dimensional eigenmatrix G M4 And the activated G M4 .

[0120] Similarly, in the right-hand branch, the antenna gain G is... t-1 With sensor data C mt The transposed layer and convolutional layer, which have a similar structure to the left branch, sequentially produce a matrix G with dimensions (N+K+N*M+9)*(N+K+N*M+9). M5 The (N+K)*(N+K) dimensional eigenmatrix G M6 And the Tanh activation function is used for G M6 Activate it. Finally, activate G. M4 and G M6 Multiplication, where G is activated M4 and G M6 These respectively characterize whether the location is used for memory updates and the specific input magnetic communication information corresponding to that location.

[0121] Finally, the product is multiplied by the output of the forget gate (N+K)*(N+K) dimension to gain the memory cell S. t1 Perform element-wise summation and map it through a fully connected layer to a gain memory cell S of dimension 1*(N+K). t At this time, the gain memory cell S t The information has been updated, including both historical magnetic communication gain information and current magnetic communication gain information.

[0122] S43, Obtain the optimal antenna gain G from the previous time step t-1. t-1 And the input C of the time-series augmentation network at the current time t. mt The input is fed into the first branch of the gain gate for concatenation, resulting in a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the eigenvector matrix G. M7 For the characteristic matrix G M7 Perform a convolution operation to obtain the convolutional feature matrix G. M8 The sigmoid activation function is used on the feature matrix G. M8 Activation is performed to obtain the feature matrix G after sigmoid activation. M8 .

[0123] The output gain of the input gate is obtained by the memory cell S. t The inverse signal query network output feature CO is input into the second branch of the gain gate for concatenation to obtain a concatenated one-dimensional vector. The concatenated one-dimensional vector is multiplied by the transpose of the concatenated one-dimensional vector to obtain the multiplied feature matrix CO1. The feature matrix CO1 is convolved to obtain the convolved feature matrix CO2. The Tanh activation function is used to activate the feature matrix CO2 to obtain the Tanh-activated feature matrix CO2.

[0124] The feature matrix G after sigmoid activation M8 Multiplying the feature matrix CO2 after Tanh activation and passing the result through a fully connected layer, we obtain the optimal antenna gain feature G at time t. t .

[0125] In one feasible implementation, the gain gate has four inputs, namely: the optimal antenna gain G at the previous moment. t-1 Its dimension is 1*(N+K); the current time is the time-to-time output of the bidirectional data conversion module's sensor data C. mt Its dimension is 1*(N*M+9); the gain memory cell S of the input gate output. t The network structure of the gain gate is as follows: The network has a dimension of 1*(N+K); the inverse signal query network outputs a feature CO with a dimension of 1*(N+K). Figure 4 As shown, in the left branch, first adjust the antenna gain G. t-1 With sensor data C mt By inputting a transpose layer, a convolutional layer, and a sigmoid activation layer with a structure similar to that of a forget gate, a matrix G of dimension (N+K+N*M+9)*(N+K+N*M+9) is obtained sequentially. M7 The (N+K)*(N+K) dimensional eigenmatrix G M7 And the activated G M7 .

[0126] In the right-hand branch, the gain memory cell S t Similar to the transposed and convolutional layers of the inverse signal query network output feature CO, the input feature CO is obtained sequentially from the (N+K)*(N+K) dimensional matrix CO1 and the (N+K)*(N+K) dimensional feature matrix CO2. The Tanh activation function is then applied to activate CO2. Finally, the activated G... M8 Multiply by CO2, where CO2 and G are activated M8 These respectively characterize the depth signal features extracted by the inverse signal query network and whether the corresponding depth signal has been introduced. Finally, the product is mapped to the optimal antenna gain feature G at the current time through a fully connected layer. t The dimension is 1*(N+K). Compared to traditional temporal neural networks, the gain gate is used to perform antenna gain feature G. t In addition to considering the inherent characteristics of time-series communication data, it also introduces deep signal features extracted by the inverse signal query network, which effectively enhances its ability to represent the gain signal at the current moment.

[0127] S5. Construct a magnetic communication enhancement network based on the bidirectional digital conversion module, the inverse signal query network, and the timing enhancement network. Obtain the output of the magnetic communication enhancement network based on the output characteristics of the inverse signal query network, the antenna gain characteristics, and the magnetic communication enhancement network.

[0128] Optionally, in S5, the magnetic communication enhancement network output is obtained based on the inverse signal query network output characteristics, antenna gain characteristics, and the magnetic communication enhancement network, including:

[0129] The inverse signal query network output feature CO and antenna gain feature G are used to determine the inverse signal query network output feature CO and antenna gain feature G. t Multiplying these matrices yields a (N+K)*(N+K) dimensional feature matrix. This feature matrix is ​​then processed through a max-pooling layer and a fully connected layer to obtain the output of a magnetic communication enhancement network with a dimension of 1*(N+K).

[0130] In the output of the magnetic communication enhancement network, each element represents the optimal array gain CG of the receiving array and the transmitting array of the communication terminal under the current pose.

[0131] In one feasible implementation, the overall network structure of the magnetic communication enhancement model is as follows: Figure 5 As shown, its construction process is as follows: Figure 6 As shown, after the sensor data is output to the bidirectional data-to-digital converter, it is input to the inverse signal query network and the timing enhancement network in different data formats. The output of the inverse signal query network is CO, and the output of the timing enhancement network is the antenna gain characteristic G. t Both have dimensions of 1*(N+K). CO is also used as part of the input to the temporal enhancement network to guide feature optimization. Next, the outputs of both are multiplied to obtain a (N+K)*(N+K) dimensional feature matrix. This matrix is ​​then processed sequentially by a max-pooling layer and a fully connected layer in the output header to obtain an output result CG with dimensions of 1*(N+K). Each element of the output result CG represents the optimal gain of the receiving array and the transmitting array of the communication terminal under the current pose. It is calculated based on the dual antenna gain prediction information from the inverse signal query network and the temporal enhancement network, resulting in more accurate prediction precision.

[0132] S6. Train the magnetic communication enhancement network based on the magnetic communication time series dataset to obtain the trained magnetic communication enhancement model.

[0133] Optionally, step S6 above may include:

[0134] The inverse signal query network is trained using the magnetic communication time series dataset to obtain the trained inverse signal query network.

[0135] The internal parameters of the pre-trained inverse signal query network are fixed, and the time-series augmentation network is trained based on the magnetic communication time-series dataset to obtain the pre-trained time-series augmentation network.

[0136] The trained inverse signal query network and the trained time series enhancement network are used as pre-trained models. The magnetic communication enhancement network is trained based on the magnetic communication time series dataset to obtain the trained magnetic communication enhancement model.

[0137] In one feasible implementation, step S6 above is the distributed generation of the magnetic communication enhancement model. First, the inverse signal query network is trained separately based on the magnetic communication time-series dataset to obtain the corresponding model; then, the internal parameters of the inverse signal query network are fixed, and the time-series enhancement network is trained to obtain the corresponding model; finally, the parameter fixing is removed, and the training model results of the above two steps are used as the pre-trained model. The magnetic communication enhancement network is trained as a whole based on the magnetic communication time-series dataset. After the training converges, the final magnetic communication enhancement model is obtained.

[0138] S7. Obtain the pose data of the magnetic communication enhancement to be performed, input the pose data into the trained magnetic communication enhancement model, and obtain the optimal array gain of the receiving array and transmitting array of the communication terminal under the current pose.

[0139] In this embodiment of the invention, the magnetic communication enhancement method based on motion posture and interference identification has a core technology of magnetic communication enhancement network. Compared with other traditional methods, the calculation process is real-time and accurate, which can significantly reduce the influence of target motion posture and surrounding interference signals. Furthermore, it can adaptively adjust the gain of different transceiver antennas based on the learning results to improve communication quality.

[0140] This magnetic communication enhancement network is a dual-branch composite neural network. The inverse signal query network has the powerful feature fitting ability of convolutional structures, and the temporal enhancement network is suitable for capturing long-term dependencies in temporal communication signals. Furthermore, the temporal enhancement network innovatively introduces convolutional features to guide the convergence of temporal features, realizing information interaction and deep fusion of dual-branch networks. Compared with single network structures or independent dual-branch structures, this structure can more comprehensively and accurately explore the magnetic sensing mechanism, thereby improving the accuracy of optimal antenna gain evaluation.

[0141] Figure 7 This is a block diagram illustrating a magnetic communication enhancement device based on motion posture and interference identification, according to an exemplary embodiment. The device is used in a magnetic communication enhancement method based on motion posture and interference identification. (Refer to...) Figure 7 The device includes a magnetic communication time-series dataset construction module 310, a bidirectional data conversion module 320, an inverse signal query network module 330, a time-series enhancement network module 340, a magnetic communication enhancement network module 350, a training module 360, and an output module 370. Among them:

[0142] The magnetic communication timing dataset construction module 310 is used to construct the magnetic communication timing dataset; wherein, the magnetic communication timing dataset includes IMU output, receiving array channel quality parameters and optimal array gain under different poses.

[0143] The bidirectional data conversion module 320 is used to input the magnetic communication time series dataset into the bidirectional data conversion module to obtain the input of the inverse signal query network and the input of the time series enhancement network.

[0144] The inverse signal query network module 330 is used to design an inverse signal query network based on the RepViT network, and to obtain the output features of the inverse signal query network based on the input and output features of the inverse signal query network.

[0145] The timing enhancement network module 340 is used to obtain antenna gain characteristics based on the input of the timing enhancement network, the output characteristics of the inverse signal query network, and the timing enhancement network.

[0146] The magnetic communication enhancement network module 350 is used to construct a magnetic communication enhancement network based on the bidirectional digital conversion module, the inverse signal query network, and the timing enhancement network. The output of the magnetic communication enhancement network is obtained based on the output characteristics of the inverse signal query network, the antenna gain characteristics, and the magnetic communication enhancement network.

[0147] Training module 360 ​​is used to train the magnetic communication enhancement network based on the magnetic communication time series dataset to obtain the trained magnetic communication enhancement model.

[0148] The output module 370 is used to acquire the pose data of the magnetic communication enhancement to be performed, input the pose data into the trained magnetic communication enhancement model, and obtain the optimal array gain of the receiving array and transmitting array of the communication terminal under the current pose.

[0149] In this embodiment of the invention, the magnetic communication enhancement method based on motion posture and interference identification has a core technology of magnetic communication enhancement network. Compared with other traditional methods, the calculation process is real-time and accurate, which can significantly reduce the influence of target motion posture and surrounding interference signals. Furthermore, it can adaptively adjust the gain of different transceiver antennas based on the learning results to improve communication quality.

[0150] This magnetic communication enhancement network is a dual-branch composite neural network. The inverse signal query network has the powerful feature fitting ability of convolutional structures, and the temporal enhancement network is suitable for capturing long-term dependencies in temporal communication signals. Furthermore, the temporal enhancement network innovatively introduces convolutional features to guide the convergence of temporal features, realizing information interaction and deep fusion of dual-branch networks. Compared with single network structures or independent dual-branch structures, this structure can more comprehensively and accurately explore the magnetic sensing mechanism, thereby improving the accuracy of optimal antenna gain evaluation.

[0151] Figure 8 This is a schematic diagram of the structure of a magnetic communication enhancement device provided in an embodiment of the present invention, as shown below. Figure 8 As shown, the magnetic communication enhancement device may include the above-mentioned Figure 7 The illustrated magnetic communication enhancement device is based on motion posture and interference recognition. Optionally, the magnetic communication enhancement device 410 may include a first processor 2001.

[0152] Optionally, the magnetic communication enhancement device 410 may also include a memory 2002 and a transceiver 2003.

[0153] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0154] The following is combined Figure 8 A detailed description of each component of the magnetic communication enhancement device 410 is provided below:

[0155] The first processor 2001 is the control center of the magnetic communication enhancement device 410. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0156] Optionally, the first processor 2001 can perform various functions of the magnetic communication enhancement device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0157] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 8 CPU0 and CPU1 are shown in the diagram.

[0158] In a specific implementation, as one example, the magnetic communication enhancement device 410 may also include multiple processors, for example... Figure 8 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). Here, "processor" can refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0159] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0160] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be connected via the interface circuit of the magnetic communication enhancement device 410. Figure 8 (Not shown in the figure) is coupled to the first processor 2001, and the embodiments of the present invention do not specifically limit this.

[0161] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0162] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 8 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the sending function.

[0163] Optionally, the transceiver 2003 can be integrated with the first processor 2001, or it can exist independently and be connected via the interface circuit of the magnetic communication enhancement device 410. Figure 8 (Not shown in the figure) is coupled to the first processor 2001, and the embodiments of the present invention do not specifically limit this.

[0164] It should be noted that, Figure 8 The structure of the magnetic communication enhancement device 410 shown does not constitute a limitation on the router. Actual knowledge structure identification devices may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0165] Furthermore, the technical effects of the magnetic communication enhancement device 410 can be referred to the technical effects of the magnetic communication enhancement method based on motion posture and interference identification in the above method embodiments, which will not be repeated here.

[0166] It should be understood that the first processor 2001 in this embodiment of the invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0167] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0168] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0169] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0170] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0171] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0172] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0173] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0174] In the embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0175] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0176] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0177] If the functionality is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0178] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A magnetic communication enhancement method based on motion posture and interference recognition, characterized in that, The method includes: S1. Construct a magnetic communication time series dataset; wherein, the magnetic communication time series dataset includes IMU output, receiving array channel quality parameters and optimal array gain under different poses; S2. Input the magnetic communication time series dataset into the bidirectional data conversion module to obtain the input of the inverse signal query network and the input of the time series enhancement network; S3. Design an inverse signal query network based on the RepViT network, and obtain the output features of the inverse signal query network according to the input of the inverse signal query network and the inverse signal query network. S4. Based on the input of the timing enhancement network, the output characteristics of the inverse signal query network, and the timing enhancement network, the antenna gain characteristics are obtained. S5. Construct a magnetic communication enhancement network based on the bidirectional digital conversion module, the inverse signal query network, and the timing enhancement network. Obtain the output of the magnetic communication enhancement network based on the output characteristics of the inverse signal query network, the antenna gain characteristics, and the magnetic communication enhancement network. S6. Train the magnetic communication enhancement network based on the magnetic communication time series dataset to obtain a trained magnetic communication enhancement model; S7. Obtain the pose data of the magnetic communication enhancement to be performed, input the pose data into the trained magnetic communication enhancement model, and obtain the optimal array gain of the receiving array and transmitting array of the communication terminal under the current pose.

2. The magnetic communication enhancement method based on motion posture and interference recognition according to claim 1, characterized in that, The construction of the magnetic communication time series dataset in S1 includes: Acquire the IMU output, receiver array channel quality parameters, and optimal array gain of the receiver and transmitter arrays of the communication terminal when the terminal moves in different poses and along different acquisition paths. The communication terminal carries a transmitting array, a receiving array, and an IMU; Different movement segments in the acquisition path introduce various types of interference signals; The data dimension of the IMU output is 1*9; the data dimension of the receiving array channel quality parameters is N*M, where N represents the number of sensors and M is the number of channel parameters for each sensor; the data dimension of the optimal array gain is N+K, where K is the number of transmit antennas.

3. The magnetic communication enhancement method based on motion posture and interference recognition according to claim 1, characterized in that, The step S2, which involves inputting the magnetic communication time-series dataset into the bidirectional data conversion module to obtain the inputs to the inverse signal query network and the time-series enhancement network, includes: The channel quality parameters of the receiving array are expanded to obtain a 1*(N*M) dimensional expanded vector. This expanded vector is then concatenated with the IMU output to obtain the input C of the 1*(N*M+9) dimensional timing enhancement network. mt Where N represents the number of sensors, and M is the number of channel parameters for each sensor; The elements in the IMU output are multiplied element by element with the channel quality parameters of the receiving array to obtain 9 matrix channels. The 9 matrix channels are then concatenated to obtain the input C of the N*M*9 dimensional inverse signal query network.

4. The magnetic communication enhancement method based on motion posture and interference recognition according to claim 1, characterized in that, The S3 section describes the design of an inverse signal query network based on the RepViT network. Based on the input and output of the inverse signal query network, the characteristics of the inverse signal query network output are obtained, including: The input C of the inverse signal query network is fed into a convolutional layer with a 1*1 kernel to obtain an N*M*3 dimensional output matrix; where N represents the number of sensors and M is the number of channel parameters for each sensor. The N*M*3 dimensional output matrix is ​​input into the RepViT network for magnetic signal feature extraction, resulting in the output feature CO of the inverse signal query network with dimension 1*(N+K); where K is the number of transmitting antennas.

5. The magnetic communication enhancement method based on motion posture and interference recognition according to claim 1, characterized in that, The temporal augmentation network includes a forget gate, an input gate, and a gain gate; The antenna gain characteristics obtained in S4 based on the input of the timing enhancement network, the output characteristics of the inverse signal query network, and the timing enhancement network include: S41. Obtain the optimal antenna gain G from the previous time t-1. t-1 And the input C of the time-series augmentation network at the current time t. mt The concatenation is performed in the transpose layer of the forget gate to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose matrix to obtain the multiplied feature matrix G. M1 For the feature matrix G M1 Perform a convolution operation to obtain the convolutional feature matrix G. M2 The sigmoid activation function is applied to the feature matrix G. M2 Activation is performed to obtain the feature matrix G after sigmoid activation. M2 The feature matrix G activated by sigmoid M2 With gain memory cells S t-1 Multiplying these values ​​yields the output gain of the forgetting gate, S, for the memory cell. t1 ; S42, Obtain the optimal antenna gain G from the previous time step t-1. t-1 And the input C of the time-series augmentation network at the current time t. mt The input gate's first branch concatenates the vectors to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the resulting feature matrix G. M3 For the feature matrix G M3 Perform a convolution operation to obtain the convolutional feature matrix G. M4 The sigmoid activation function is applied to the feature matrix G. M4 Activation is performed to obtain the feature matrix G after sigmoid activation. M4 ; The optimal antenna gain G obtained at the previous time t-1 t-1 And the input C of the time-series augmentation network at the current time t. mt In the second branch of the input gate, concatenation is performed to obtain a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the resulting feature matrix G. M5 For the feature matrix G M5 Perform a convolution operation to obtain the convolutional feature matrix G. M6 The Tanh activation function is applied to the feature matrix G. M6 Activation is performed to obtain the feature matrix G after Tanh activation. M6 ; The sigmoid-activated feature matrix G M4 With the Tanh-activated feature matrix G M6 Multiply the result and then multiply it with the output gain of the forget gate for the memory cell S. t1 Element-wise summation is performed, and the summation result is passed through a fully connected layer to obtain the output gain of the input gate, memory cell S. t ; S43, Obtain the optimal antenna gain G from the previous time step t-1. t-1 And the input C of the time-series augmentation network at the current time t. mt The input is fed into the first branch of the gain gate for concatenation, resulting in a concatenated one-dimensional vector. This concatenated one-dimensional vector is then multiplied by its transpose to obtain the eigenvector matrix G. M7 For the feature matrix G M7 Perform a convolution operation to obtain the convolutional feature matrix G. M8 The sigmoid activation function is applied to the feature matrix G. M8 Activation is performed to obtain the feature matrix G after sigmoid activation. M8 ; The output gain of the input gate is obtained by the memory cell S. t The inverse signal query network output feature CO is input into the second branch of the gain gate for concatenation to obtain a concatenated one-dimensional vector. The concatenated one-dimensional vector is multiplied by the transpose of the concatenated one-dimensional vector to obtain the multiplied feature matrix CO1. The feature matrix CO1 is convolved to obtain the convolved feature matrix CO2. The Tanh activation function is used to activate the feature matrix CO2 to obtain the Tanh-activated feature matrix CO2. The sigmoid-activated feature matrix G M8 Multiplying the Tanh-activated feature matrix CO2 by the multiplication result and passing it through a fully connected layer, we obtain the optimal antenna gain feature G at the current time t. t .

6. The magnetic communication enhancement method based on motion posture and interference recognition according to claim 1, characterized in that, The step S5, which involves obtaining the magnetic communication enhancement network output based on the inverse signal query network output characteristics, the antenna gain characteristics, and the magnetic communication enhancement network, includes: The output feature CO of the inverse signal query network and the antenna gain feature G are used. t Multiply to obtain a (N+K)*(N+K) dimensional feature matrix. Process the feature matrix through a max pooling layer and a fully connected layer to obtain a magnetic communication enhancement network output with a dimension of 1*(N+K). In the output of the magnetic communication enhancement network, each element represents the optimal array gain CG of the receiving array and the transmitting array of the communication terminal under the current pose.

7. The magnetic communication enhancement method based on motion posture and interference recognition according to claim 1, characterized in that, The step S6 involves training the magnetic communication enhancement network based on the magnetic communication time-series dataset to obtain a trained magnetic communication enhancement model, including: The inverse signal query network is trained based on the magnetic communication time series dataset to obtain the trained inverse signal query network; The internal parameters of the trained inverse signal query network are fixed, and the time-series enhancement network is trained based on the magnetic communication time-series dataset to obtain the trained time-series enhancement network. The trained inverse signal query network and the trained time series enhancement network are used as pre-trained models. The magnetic communication enhancement network is trained based on the magnetic communication time series dataset to obtain the trained magnetic communication enhancement model.

8. A magnetic communication enhancement device based on motion posture and interference recognition, wherein the magnetic communication enhancement device based on motion posture and interference recognition is used to implement the magnetic communication enhancement method based on motion posture and interference recognition as described in any one of claims 1-7, characterized in that, The device includes: A magnetic communication timing dataset construction module is used to construct a magnetic communication timing dataset; wherein, the magnetic communication timing dataset includes IMU output, receiver array channel quality parameters and optimal array gain under different poses; A bidirectional data conversion module is used to input the magnetic communication time series dataset into the bidirectional data conversion module to obtain the input of the inverse signal query network and the input of the time series enhancement network; The inverse signal query network module is used to design an inverse signal query network based on the RepViT network, and to obtain the output features of the inverse signal query network based on the input of the inverse signal query network and the inverse signal query network itself. The timing enhancement network module is used to obtain antenna gain characteristics based on the input of the timing enhancement network, the output characteristics of the inverse signal query network, and the timing enhancement network itself. The magnetic communication enhancement network module is used to construct a magnetic communication enhancement network based on the bidirectional digital conversion module, the inverse signal query network, and the timing enhancement network, and to obtain the output of the magnetic communication enhancement network based on the output characteristics of the inverse signal query network, the antenna gain characteristics, and the magnetic communication enhancement network. The training module is used to train the magnetic communication enhancement network based on the magnetic communication time series dataset to obtain a trained magnetic communication enhancement model. The output module is used to acquire the pose data of the magnetic communication enhancement to be performed, input the pose data into the trained magnetic communication enhancement model, and obtain the optimal array gain of the receiving array and transmitting array of the communication terminal under the current pose.

9. A magnetic communication enhancement device, characterized in that, The magnetic communication enhancement device includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 7.

Images