Positioning method, electronic equipment, program product and storage medium

Abstract

The invention discloses a positioning method, electronic equipment, a program product and a storage medium. The positioning method comprises the following steps: acquiring a receiving signal received by an access point, and dividing a sensing area of the access point into a plurality of sub-areas; preprocessing the received signal by using a preset signal model to obtain a to-be-predicted signal; determining a target sub-region corresponding to the to-be-predicted signal from the sensing region according to the to-be-predicted signal and a pre-trained region selection network; and determining a target position prediction model corresponding to the target sub-regions, inputting the to-be-predicted signal to the target position prediction model, and predicting to obtain position data, the plurality of sub-regions being in one-to-one correspondence with a plurality of position prediction models. According to the invention, the technical problems of complex data acquisition and large calculated amount of a positioning method in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of signal processing technology, and in particular to a positioning method, electronic equipment, program product and storage medium. Background Art

[0002] The non-cellular massive multiple-input multiple-output (Massive MIMO) system is widely used in 5G networks. Its performance advantage lies in that under the same frequency resources and power conditions, it can provide much more channel capacity than the traditional multiple-input multiple-output (MIMO) system, thereby supporting more terminal access and higher service rates.

[0003] Large-scale MIMO systems mainly rely on multi-antenna array receivers to coherently process multipath signals to obtain the advantages of frequency selectivity, spatial selectivity, and angle selectivity, and then use multi-antenna array receivers to perform spatial multiplexing and multiple access to multiple users. In sensor networks, target positioning is usually achieved through ranging and angle measurement. Despite attempts to address practical challenges in target positioning, factors such as imperfect time synchronization, sensor position disturbances, and gain phase mismatches in multi-channel receivers hinder further improvements in positioning accuracy. To overcome these challenges, deep learning has powerful nonlinear fitting capabilities and points out a new paradigm for achieving high-precision positioning. However, deep learning requires pre-training of large-scale data in the offline stage, and its application in a wide range of perception areas will lead to complex data collection and huge computational burden.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a positioning method, an electronic device, a program product and a storage medium, so as to at least solve the technical problems of complex data collection and large amount of calculation in the positioning method in the prior art.

[0006] According to one aspect of an embodiment of the present invention, a positioning method is provided, including: acquiring a received signal received by an access point, wherein a perception area of ​​the access point is divided into multiple sub-areas; using a preset signal model, preprocessing the received signal to obtain a signal to be predicted; determining a target sub-area corresponding to the signal to be predicted from the perception area according to the signal to be predicted and a pre-trained area selection network; determining a target position prediction model corresponding to the target sub-area, and inputting the signal to be predicted into the target position prediction model to predict position data, wherein multiple sub-areas correspond one-to-one to multiple position prediction models.

[0007] Optionally, constructing the preset signal model includes: determining a communication signal expression based on user equipment configuration information and preset access point configuration information of a receiving access point, wherein the user equipment is included in a non-cellular multiple-input multiple-output network; determining a perception signal expression based on preset access point configuration information of a sending access point and a receiving access point, wherein the sending access point and the receiving access point are included in the non-cellular multiple-input multiple-output network; determining the preset signal model based on the communication signal expression, the perception signal expression and additive noise.

[0008] Optionally, determining a communication signal expression based on user equipment configuration information and preset access point configuration information of a receiving access point includes: determining a communication signal expression based on a communication channel representation and a communication signal representation of the user equipment and the receiving access point, wherein the communication channel representation and the communication signal representation are included in the user equipment configuration information and the preset access point configuration information.

[0009] Optionally, determining the perception signal expression based on preset configuration information of the sending access point and the receiving access point includes: determining the perception signal expression based on the perception channel representation and the perception signal representation of the sending access point and the receiving access point, wherein the perception channel representation and the perception signal representation are included in the preset access point configuration information.

[0010] Optionally, the perceived channel representation includes: signal attenuation of the transmitting access point and the receiving access point, the dual-base distance between the transmitting access point and the receiving access point, the steering vectors of the transmitting access point and the receiving access point, the subcarrier spacing, the speed of light, the coherent processing interval, the Doppler frequency shift, and the carrier wavelength.

[0011] Optionally, a preset signal model is used to preprocess the received signal to obtain a signal to be predicted, including: based on the preset signal model, determining multiple perception signals from multiple received signals; determining a signal matrix based on the multiple perception signals, wherein the signal matrix is ​​a complex matrix; dividing the real part and the imaginary part in the signal matrix to determine the signal to be predicted.

[0012] Optionally, based on the signal to be predicted and a pre-trained region selection network, a target sub-region corresponding to the signal to be predicted is determined from the perception area, including: inputting the signal to be predicted into a feature extraction module in the region selection network to perform feature extraction and obtain a feature extraction result; inputting the feature extraction result into a regression module in the region selection network to determine the target sub-region.

[0013] Optionally, a target position prediction model corresponding to the target sub-region is determined, and the signal to be predicted is input into the target position prediction model to predict the position data, including: determining the target position prediction model from multiple position prediction models according to the target sub-region; inputting the signal to be predicted into a spatial attention module in the target position prediction model, and using the spatial attention module to perform weighted processing on the signal to be predicted to obtain weighted features; and inputting the weighted features into a residual convolutional neural network in the target position prediction model for prediction to obtain position data.

[0014] Optionally, the spatial attention module includes a pooling layer, a convolution layer and an activation function. The signal to be predicted is input into the spatial attention module in the target position prediction model, and the spatial attention module is used to perform weighted processing on the signal to be predicted to obtain a weighted feature, including: using the pooling layer to perform pooling processing on the signal to be predicted to obtain a pooling processing result; using the convolution layer to perform a convolution operation on the pooling processing result to obtain a convolution processing result; using the activation function to weight limit the convolution processing result based on a preset weight range to obtain the spatial attention weight; element-wise multiplying the spatial attention weight with the signal to be predicted to obtain a weighted feature.

[0015] Optionally, the target position prediction model includes a first convolution module, a second convolution module and a regression module, and the weighted features are input into the residual convolution neural network in the target position prediction model for prediction to obtain position data, including: using the first convolution module to perform a first processing on the weighted features to obtain a first processing result, wherein the first processing includes convolution processing, normalization processing, activation processing and pooling processing in sequence; using the second convolution module to perform a second processing on the first processing result to obtain a second processing result, wherein the second processing includes a main branch operation and an auxiliary branch operation; using the regression module to predict the second processing result to obtain position data.

[0016] Optionally, a second convolution module is used to perform a second processing on the first processing result to obtain a second processing result, including: using the second convolution module to perform a main branch operation on the first processing result to obtain a main branch operation result, wherein the main branch operation includes, in sequence: first convolution processing, normalization processing, activation processing, second convolution processing, and second normalization processing; using the second convolution module to perform an auxiliary branch operation on the first processing result to obtain an auxiliary branch operation result, wherein the auxiliary branch operation includes: convolution processing based on a preset dimension; determining the second processing result according to the main branch operation result and the auxiliary branch operation result.

[0017] According to another aspect of an embodiment of the present invention, an electronic device is provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes any one of the above positioning methods when running.

[0018] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the positioning method in any one of the above items is implemented.

[0019] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium including a stored executable program, wherein when the executable program runs, the device where the computer-readable storage medium is located is controlled to execute any of the above positioning methods.

[0020] In an embodiment of the present invention, a received signal received by an access point is obtained, wherein the perception area of ​​the access point is divided into a plurality of sub-areas; the received signal is preprocessed using a preset signal model to obtain a signal to be predicted; a target sub-area corresponding to the signal to be predicted is determined from the perception area according to the signal to be predicted and a pre-trained area selection network; a target position prediction model corresponding to the target sub-area is determined, and the signal to be predicted is input into the target position prediction model to predict position data, wherein a plurality of sub-areas correspond one to one to a plurality of position prediction models. The present invention divides the perception area into a plurality of areas through an area selection network, and different areas are positioned using different positioning models, which can solve the problem of data collection and computational burden caused by applying deep learning to achieve target positioning in a wide range of perception areas, thereby solving the technical problems of complex data collection and large amount of computation in the positioning method in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0022] Figure 1 is a flow chart of a positioning method according to one embodiment of the present invention;

[0023] Figure 2 is a schematic diagram of a non-cellular massive MIMO system according to one embodiment of the present invention;

[0024] Figure 3 is a schematic diagram of data preprocessing according to one embodiment of the present invention;

[0025] Figure 4 is a schematic diagram of a region selection network according to one embodiment of the present invention;

[0026] Figure 5 is a schematic diagram of a spatial attention mechanism according to one embodiment of the present invention;

[0027] Figure 6is a schematic diagram of a residual convolutional neural network according to one embodiment of the present invention;

[0028] Figure 7 is a flowchart of a positioning method according to one embodiment of the present invention;

[0029] Figure 8 is a schematic diagram of performance comparison of a region selection network at different signal-to-noise ratios according to one embodiment of the present invention;

[0030] Fig. 9 is a schematic diagram of performance comparison of a region selection network at different grid intervals according to one embodiment of the present invention;

[0031] Fig.10 is a schematic diagram of RMSE comparison between a positioning method according to one embodiment of the present invention and other methods at different signal-to-noise ratios;

[0032] Fig.11 is a schematic diagram comparing CDFs of estimation errors of a positioning method according to one embodiment of the present invention and other methods;

[0033] Fig.12 is a structural block diagram of a collaborative perception system according to one embodiment of the present invention. DETAILED DESCRIPTION

[0034] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0035] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0036] According to an embodiment of the present invention, an embodiment of a positioning method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system containing at least one set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0037] The method embodiment can also be executed in an electronic device including a memory and a processor, a similar control device or a cloud. Taking an electronic device as an example, the electronic device may include one or more processors and a memory for storing data. Optionally, the electronic device may also include a communication device and a display device for communication functions. It can be understood by those of ordinary skill in the art that the above structural description is only illustrative and does not limit the structure of the above electronic device. For example, the electronic device may also include more or fewer components than the above structural description, or have a configuration different from the above structural description.

[0038] The processor may include one or more processing units. For example, the processor may include a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processing (DSP) chip, a microcontroller unit (MCU), a field-programmable gate array (FPGA), a neural network processor (NPU), a tensor processing unit (TPU), an artificial intelligence (AI) type processor, and the like. Among them, different processing units may be independent components or integrated into one or more processors. In some instances, the electronic device may also include one or more processors.

[0039] The memory can be used to store computer programs, such as storing the computer program corresponding to the positioning method in the embodiment of the present invention, and the processor implements the above positioning method by running the computer program stored in the memory. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory may further include a memory remotely arranged relative to the processor, and these remote memories can be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0040] The communication device is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the communication device includes a network adapter (network interface controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the communication device can be a radio frequency (radio frequency, RF) module, which is used to communicate with the Internet wirelessly. In some embodiments of the present solution, the communication device is used to connect to mobile devices such as mobile phones and tablets, and instructions can be sent to electronic devices through the mobile device.

[0041] The display device may be a touch screen type liquid crystal display (LCD) and a touch display (also referred to as a "touch screen" or "touch display screen"). The liquid crystal display may enable a user to interact with a user interface of the electronic device. In some embodiments, the electronic device has a graphical user interface (GUI), and a user may interact with the GUI by touching a finger release and / or gestures on a touch-sensitive surface, and executable instructions for executing the above-mentioned human-computer interaction functions are configured / stored in a computer program product or a readable storage medium executable by one or more processors.

[0042] Figure 1 is a flow chart of a positioning method according to one embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0043] Step S101: acquiring a received signal received by an access point, wherein a sensing area of ​​the access point is divided into a plurality of sub-areas.

[0044] Specifically, in wireless communications, access points (APs) refer to devices that can connect wireless devices to the network, such as Wi-Fi routers. Received signals refer to wireless signals that can be detected by access points, which can come from mobile devices, sensors, etc. The sensing area refers to the geographical range in which the access point can sense wireless signals. In order to improve positioning accuracy, the sensing area is divided into smaller sub-areas, each with its own characteristics and boundaries.

[0045] Step S102: preprocess the received signal using a preset signal model to obtain a signal to be predicted.

[0046] Specifically, the preset signal model is a mathematical model preset according to the physical environment and wireless propagation characteristics, which is used to describe the propagation mode of the signal in space. Preprocessing refers to the necessary mathematical processing of the received signal to eliminate noise, enhance signal characteristics, etc., so that it is more suitable for subsequent position prediction. The signal to be predicted after preprocessing is used for the next step of position prediction.

[0047] Step S103, determining a target sub-region corresponding to the signal to be predicted from the perception region according to the signal to be predicted and the pre-trained region selection network.

[0048] Specifically, the region selection network is a pre-trained machine learning model that can identify the sub-region from which the signal comes based on the signal characteristics of the input signal to be predicted. After analysis by the region selection network, the specific sub-region from which the signal to be predicted comes is determined and used as the target sub-region.

[0049] Step S104, determining the target position prediction model corresponding to the target sub-region, and inputting the signal to be predicted into the target position prediction model to predict the position data, wherein multiple sub-regions correspond one-to-one to multiple position prediction models.

[0050] Specifically, the target location prediction model is the location prediction model corresponding to the target sub-region. Each sub-region has a corresponding location prediction model, which is trained based on the wireless signal propagation characteristics of different sub-regions. The signal to be predicted is input into the location prediction model corresponding to the target sub-region, and the model will output a location prediction result, i.e., location data, which is obtained based on the signal characteristics and sub-region characteristics.

[0051] It is understandable that each sub-region has its own exclusive location prediction model, which can more accurately predict the location based on signal characteristics.

[0052] In an embodiment of the present invention, a received signal received by an access point is obtained, wherein the perception area of ​​the access point is divided into a plurality of sub-areas; the received signal is preprocessed using a preset signal model to obtain a signal to be predicted; a target sub-area corresponding to the signal to be predicted is determined from the perception area according to the signal to be predicted and a pre-trained area selection network; a target position prediction model corresponding to the target sub-area is determined, and the signal to be predicted is input into the target position prediction model to predict position data, wherein a plurality of sub-areas correspond one to one to a plurality of position prediction models. The present invention divides the perception area into a plurality of areas through an area selection network, and different areas are positioned using different positioning models, which can solve the problem of data collection and computational burden caused by applying deep learning to achieve target positioning in a wide range of perception areas, thereby solving the technical problems of complex data collection and large amount of computation in the positioning method in the prior art.

[0053] Optionally, constructing the preset signal model includes: determining a communication signal expression based on user equipment configuration information and preset access point configuration information of a receiving access point, wherein the user equipment is included in a non-cellular multiple-input multiple-output network; determining a perception signal expression based on preset access point configuration information of a sending access point and a receiving access point, wherein the sending access point and the receiving access point are included in the non-cellular multiple-input multiple-output network; determining the preset signal model based on the communication signal expression, the perception signal expression and additive noise.

[0054] Specifically, Figure 2 As shown in the figure, consider a non-cellular massive multiple-input multiple-output (MIMO) system consisting of U single-antenna user equipment (UE) and N randomly distributed access points (AP), where each AP is equipped with a row of M x Antenna and each column M y The total number of antennas per AP is M = M x M y Assume that T APs send perception signals and R APs receive the perception signals and the communication signals from UEs at the same time. Then the received signal of the rth receiving AP at the kth subcarrier and the lth symbol is for:

[0055]

[0056] in, represents the communication channel of the u-th UE, is the sensed channel from the tth transmitting AP to the rth receiving AP, s u(k,l) and They represent the communication signal from the u-th UE and the perception signal from the t-th sending AP, respectively. is the additive noise that follows a complex Gaussian distribution.

[0057] It should be noted that the above formula (1) is a preset signal model, which includes a communication signal expression, a perception signal expression and additive noise.

[0058] Optionally, determining a communication signal expression based on user equipment configuration information and preset access point configuration information of a receiving access point includes: determining a communication signal expression based on a communication channel representation and a communication signal representation of the user equipment and the receiving access point, wherein the communication channel representation and the communication signal representation are included in the user equipment configuration information and the preset access point configuration information.

[0059] Specifically, the communication signal expression is:

[0060] Optionally, determining the perception signal expression based on preset configuration information of the sending access point and the receiving access point includes: determining the perception signal expression based on the perception channel representation and the perception signal representation of the sending access point and the receiving access point, wherein the perception channel representation and the perception signal representation are included in the preset access point configuration information.

[0061] Specifically, the perception signal expression is:

[0062] Optionally, the sensing signal and the communication signal are separated by time division and orthogonal pilot signals are applied so that the receiving AP can distinguish the sensing signals from different transmitting APs. At the same time, in order to improve the directionality of the received signal, a receiving beamformer is used on the receiving AP. Therefore, the sensing signal from the rth receiving AP of the tth transmitting AP is:

[0063]

[0064] Among them, w H represents the conjugate transpose operation, is additive white Gaussian noise.

[0065] Optionally, the perceived channel representation includes: signal attenuation of the transmitting access point and the receiving access point, the dual-base distance between the transmitting access point and the receiving access point, the steering vectors of the transmitting access point and the receiving access point, the subcarrier spacing, the speed of light, the coherent processing interval, the Doppler frequency shift, and the carrier wavelength.

[0066] Specifically, the sensing channel is expressed as:

[0067]

[0068] Among them, α r,t represents the attenuation from the tth transmitting AP to the rth receiving AP, Δf represents the subcarrier spacing, c represents the speed of light, and T p represents the coherent processing interval, f d represents the Doppler frequency shift, λ represents the carrier wavelength, and its numerical calculation formula is: where f c is the subcarrier frequency, d r,t It represents the bistatic distance between the tth transmitting AP and the rth receiving AP, and its numerical calculation formula is d r +d t , where d r and d t They represent the distance between the target and the rth receiving AP and the distance between the target and the tth sending AP, respectively. Specifically, d r and d t The calculation formula is as follows:

[0069]

[0070] Among them, (x q ,y q )、(x r ,y r ) and (x t ,y t ) represent the coordinates of the target, the rth receiving AP, and the tth sending AP, respectively. and They represent the steering vectors of the rth receiving AP and the tth sending AP respectively, and are specifically expressed as follows:

[0071]

[0072] Accompanied by,

[0073]

[0074] Among them, d, θ r , and θ t denote the antenna spacing of UPA, the azimuth angle of the rth receiving AP, the elevation angle of the rth receiving AP, the azimuth angle of the tth transmitting AP, and the elevation angle of the tth transmitting AP, respectively. (·) T and denote the transpose operation and the Kronecker product, respectively.

[0075] Optionally, a preset signal model is used to preprocess the received signal to obtain a signal to be predicted, including: based on the preset signal model, determining multiple perception signals from multiple received signals; determining a signal matrix based on the multiple perception signals, wherein the signal matrix is ​​a complex matrix; dividing the real part and the imaginary part in the signal matrix to determine the signal to be predicted.

[0076] For example, the sensing area is divided along the XY plane into A={A1, A2, ..., A S}, where A represents a set of sub-regions, and each element represents a divided sub-region. Figure 3 As shown, it is assumed that the perception signal determined from the received signals of the t-th transmitting AP and the r-th receiving AP is Stack all signals to form a signal matrix KL and TR are the product of the subcarrier and symbol and the product of the number of sending APs and receiving APs respectively. Signal Matrix is a complex matrix, but the input of deep learning algorithms must be real numbers. To this end, the signal matrix The real and imaginary parts of are divided into two channels of the input features, expressed as This is the signal to be predicted.

[0077] Optionally, based on the signal to be predicted and a pre-trained region selection network, a target sub-region corresponding to the signal to be predicted is determined from the perception area, including: inputting the signal to be predicted into a feature extraction module in the region selection network to perform feature extraction and obtain a feature extraction result; inputting the feature extraction result into a regression module in the region selection network to determine the target sub-region.

[0078] Specifically, Figure 4 As shown in Figure 1, the region selection network aims to identify the sub-region to which the target belongs, including a feature extraction module and a regression module. The feature extraction module includes three convolutional layers, each of which includes a convolution operation, batch normalization, an activation function, and a maximum pooling. The regression module includes a global average pooling and a fully connected layer, and the output label is The element corresponding to the sub-region to which the target belongs is set to 1, and the remaining elements are set to 0.

[0079] Optionally, a target position prediction model corresponding to the target sub-region is determined, and the signal to be predicted is input into the target position prediction model to predict the position data, including: determining the target position prediction model from multiple position prediction models according to the target sub-region; inputting the signal to be predicted into a spatial attention module in the target position prediction model, and using the spatial attention module to perform weighted processing on the signal to be predicted to obtain weighted features; and inputting the weighted features into a residual convolutional neural network in the target position prediction model for prediction to obtain position data.

[0080] Specifically, Figure 5 As shown in Figure 2, the spatial attention mechanism is applied to give appropriate contribution weights to target localization to different transceiver stations. The output of the maximum pooling and average pooling is used as the input of the convolutional layer, and then the Sigmoid activation function is applied to limit the attention weights between 0 and 1, which is similar to Perform element-wise multiplication to obtain weighted features.

[0081] Finally, the weighted features output by the spatial attention mechanism are used as the input of the residual convolutional neural network to output the location coordinates of the target. Figure 6 As shown in Figure 1, the residual convolutional neural network includes a basic convolution module, a refined convolution module, and a regression module. The basic convolution module consists of a convolution layer, including convolution operation, batch normalization, activation function, and maximum pooling. The refined convolution module contains three residual blocks. The input features of each residual block undergo a sequential process, first through the convolution operation, followed by batch normalization and ReLU activation function, and then another convolution operation and batch normalization are applied, which is called the main branch operation. There are two methods for the auxiliary branch operation. If the stride is 1, the input and output features of the residual block have the same dimension, allowing them to be directly connected. If the stride is 2, the input features of the residual block are processed by a convolution operation with a kernel size of [1,1] to obtain the same dimension as the output feature. The output of the main branch is added to the output of the auxiliary branch in an element-wise manner to represent the result of residual learning. Finally, the result of residual learning is passed through a ReLU activation function.

[0082] Optionally, the spatial attention module includes a pooling layer, a convolution layer and an activation function. The signal to be predicted is input into the spatial attention module in the target position prediction model, and the spatial attention module is used to perform weighted processing on the signal to be predicted to obtain a weighted feature, including: using the pooling layer to perform pooling processing on the signal to be predicted to obtain a pooling processing result; using the convolution layer to perform a convolution operation on the pooling processing result to obtain a convolution processing result; using the activation function to weight the convolution processing result based on a preset weight range to obtain the spatial attention weight; element-wise multiplying the spatial attention weight with the signal to be predicted to obtain a weighted feature.

[0083] Specifically, Figure 5 As shown, the signal to be predicted Through the maximum pooling layer and the average pooling layer, the outputs of the maximum pooling layer and the average pooling layer are connected and used as the input of the convolution layer. The Sigmoid activation function is further applied to limit the attention weight of the convolution layer output to between 0 and 1, and then the attention weight is combined with Perform element-wise multiplication to obtain weighted features.

[0084] Optionally, the target position prediction model includes a first convolution module, a second convolution module and a regression module, and the weighted features are input into the residual convolution neural network in the target position prediction model for prediction to obtain position data, including: using the first convolution module to perform a first processing on the weighted features to obtain a first processing result, wherein the first processing includes convolution processing, normalization processing, activation processing and pooling processing in sequence; using the second convolution module to perform a second processing on the first processing result to obtain a second processing result, wherein the second processing includes a main branch operation and an auxiliary branch operation; using the regression module to predict the second processing result to obtain position data.

[0085] Specifically, Figure 6 As shown, the basic convolution module (i.e., the first convolution module) is first used to perform a first processing on the weighted features, and then the refined convolution module (i.e., the second convolution module) is used to perform a second processing on the output of the basic convolution module, and finally the regression module is used to predict the output of the refined convolution module to obtain the position coordinates of the target, i.e., the position data.

[0086] Optionally, a second convolution module is used to perform a second processing on the first processing result to obtain a second processing result, including: using the second convolution module to perform a main branch operation on the first processing result to obtain a main branch operation result, wherein the main branch operation includes, in sequence: first convolution processing, normalization processing, activation processing, second convolution processing, and second normalization processing; using the second convolution module to perform an auxiliary branch operation on the first processing result to obtain an auxiliary branch operation result, wherein the auxiliary branch operation includes: convolution processing based on a preset dimension; determining the second processing result according to the main branch operation result and the auxiliary branch operation result.

[0087] Specifically, Figure 6 As shown in the figure, the refined convolution module (i.e., the second convolution module) contains three residual blocks. The processing process of the input features of each residual block is as follows: first, through the convolution operation (i.e., the first convolution processing), followed by batch normalization and ReLU activation function processing (i.e., normalization processing and activation processing), and then using another convolution operation and batch normalization (i.e., the second convolution processing and the second normalization processing). This process is called the main branch operation. There are two methods for the auxiliary branch operation. If the stride is 1, the input and output features of the residual block have the same dimension, allowing them to be directly connected. If the stride is 2, the input features of the residual block are processed by a convolution operation with a kernel size of [1,1] to obtain the same dimension as the output features. The output of the main branch is added to the output of the auxiliary branch in an element-wise manner to represent the result of residual learning. Finally, the result of residual learning is passed through a ReLU activation function to obtain the second processing result.

[0088] For example, Figure 7As shown, the positioning method provided by the present invention has the following process:

[0089] When target positioning is required, the received signal is first preprocessed to obtain input features Based on the trained region selection network, the sub-region to which the target belongs is determined. Then, within the determined sub-region, the position coordinates of the target are output based on the trained spatial attention mechanism and residual convolutional neural network, as follows:

[0090] Preprocess the received signal and determine the perception signal from the received signal; divide the perception area into S sub-areas along the XY plane; process the perception signal to obtain a signal matrix The signal matrix Convert to That is, the signal matrix The real and imaginary parts of are divided into two channels of the input features, expressed as Get training data and divide the training set and test set into a ratio of 8:2; set loss functions l1 and l2, batches B1 and B2, iterations E1 and E2, optimizers O1 and O2; train region selection network based on l1, B1, E1, and O1; train spatial attention mechanism and residual convolutional neural network based on l2, B2, E2, and O2; use the trained network to achieve target positioning. Specifically, through the trained region selection network, based on Determine the sub-region of the target, through the trained spatial attention mechanism and residual convolutional neural network, based on Achieve precise target positioning; output target coordinates.

[0091] Exemplarily, the positioning method provided by the present invention is simulated as follows:

[0092] First, construct the simulation experiment parameters, the parameter table is as follows Table 1:

[0093] Table 1

[0094] parameter value Carrier frequency 30GHz Subcarrier spacing 60kHz Number of subcarriers 1024 Number of symbols 2 Antenna spacing Half wavelength Send AP coordinates (4.8,1.2)，(5.2,9.2) Receive AP coordinates (2.4,0)，(7.2,0)，(2.8,10)，(7.6,10)

[0095] The simulation constructs a scenario with a coverage area of ​​10m×10m, where two sending APs and four receiving APs are deployed at specific locations, and each AP is equipped with a x =8 and M y= 4 UPA antenna arrays, the system and waveform parameters used in the simulation are shown in Table 1. In the data preprocessing stage, the large sensing area is first divided into four sub-areas of equal size. Then, the entire area is evenly divided at intervals of 0.1m, and the received signals at these points are marked as the indexes corresponding to the sub-areas, thereby constructing a data set of size 10000 for the area selection network. In addition, in order to achieve high-precision target positioning, points are uniformly selected at intervals of 0.025m in each sub-area, and the received signals at these points are marked with corresponding coordinates, thereby constructing a data set of size 4×40000 for the spatial attention mechanism and residual convolutional neural network. On the area selection network, the kernel size, stride and padding of the three convolution operations are [3,3], 1 and 2, respectively, and the number of output channels is 32, 64 and 128, respectively. The kernel size and padding of the convolution operation in the spatial attention mechanism are [1,7] and [0,3], respectively. In the residual convolutional neural network, the kernel size of all convolution operations is [3,3], and a stride of 2 is selected to connect the residual blocks. At this time, the stride of the convolution operation of the auxiliary channel is 4. The Adam optimizer is used to update the network parameters, and the cross entropy loss function and the minimum mean square error loss function are used to handle the region selection task and the target localization task respectively. The learning rate and mini-batch size are set to 0.001 and 40 respectively. In addition, the learning rate is reduced by 0.15 times every 25 rounds.

[0096] Figure 8 , Fig. 9 The effect of different signal-to-noise ratios and grid spacing on the training accuracy of the region selection network is shown. Figure 8 It can be seen that the training accuracy is positively correlated with the signal-to-noise ratio, especially in the range of 10dB to 15dB, which increases by about 5.63%. In addition, the convergence speed is not affected by the signal-to-noise ratio. At the same time, when the signal-to-noise ratio is 15dB, the training accuracy of the region selection network is 96.65% and the test accuracy is 96.52%, which has excellent classification accuracy and strong generalization ability. In addition, Fig. 9 The figure shows the effect of grid spacing on training accuracy when the signal-to-noise ratio is fixed at 15dB. Obviously, the performance improves as the grid spacing decreases. However, a smaller grid spacing means that more data needs to be obtained, which will lead to higher deployment costs. An optimal trade-off should be sought between classification accuracy and cost based on actual requirements.

[0097] To further compare the performance of target localization, the proposed method is compared with BackProjection, Transformer, and FedSLoc. Figure 10 to Figure 11 The performance comparison of different methods is shown in many aspects. It is worth noting that the proposed method is called RSAResNet. Fig.10It can be seen that the RMSE of the four methods decreases with the increase of SNR. Specifically, Back Projection performs the worst, while FedSLoc performs slightly better than Transformer. The proposed method is significantly better than the other three methods, with positioning accuracy increased by 68.2%, 49.6%, and 47.1%, respectively, which shows that the proposed method performs well in high-precision target positioning. Fig.11 The cumulative distribution function (CDF) of the estimation error of different methods is described in sub-region 1 and when the signal-to-noise ratio is 15dB. It can be seen that the other three methods all show similar positioning accuracy, but they all lag behind the proposed method significantly, which further strengthens the verification of the effectiveness of the proposed method.

[0098] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0099] In the embodiments of the present invention, a collaborative perception system is also provided to implement the above-mentioned embodiments and preferred implementation modes, which have been described and will not be repeated here. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0100] Fig.12 is a block diagram of a collaborative sensing system 200 according to one embodiment of the present invention. Fig.12As shown, the system includes: an acquisition module 201, which is used to acquire a received signal received by an access point, wherein the perception area of ​​the access point is divided into multiple sub-areas; a preprocessing module 202, which is used to preprocess the received signal using a preset signal model to obtain a signal to be predicted; a determination module 203, which is used to determine a target sub-area corresponding to the signal to be predicted from the perception area according to the signal to be predicted and a pre-trained area selection network; a prediction module 204, which is used to determine a target position prediction model corresponding to the target sub-area, and input the signal to be predicted into the target position prediction model to predict the position data, wherein multiple sub-areas correspond one-to-one to multiple position prediction models.

[0101] Optionally, the preprocessing module 202 is also used to: determine a communication signal expression based on user equipment configuration information and preset access point configuration information of a receiving access point, wherein the user equipment is included in a non-cellular multiple-input multiple-output network; determine a perception signal expression based on preset access point configuration information of a sending access point and a receiving access point, wherein the sending access point and the receiving access point are included in the non-cellular multiple-input multiple-output network; determine a preset signal model based on the communication signal expression, the perception signal expression and additive noise.

[0102] Optionally, the preprocessing module 202 is further used to determine a communication signal expression based on a communication channel representation and a communication signal representation of the user equipment and the receiving access point, wherein the communication channel representation and the communication signal representation are included in the user equipment configuration information and the preset access point configuration information.

[0103] Optionally, the pre-processing module 202 is further configured to: determine a perception signal expression based on the perception channel representation and the perception signal representation of the sending access point and the receiving access point, wherein the perception channel representation and the perception signal representation are included in the preset access point configuration information.

[0104] Optionally, the perceived channel representation determined by the pre-processing module 202 includes: signal attenuation of the transmitting access point and the receiving access point, the dual-base distance between the transmitting access point and the receiving access point, the steering vectors of the transmitting access point and the receiving access point, the subcarrier spacing, the speed of light, the coherent processing interval, the Doppler frequency shift, and the carrier wavelength.

[0105] Optionally, the preprocessing module 202 is further used to: determine multiple perception signals from multiple received signals based on a preset signal model; determine a signal matrix based on the multiple perception signals, wherein the signal matrix is ​​a complex matrix; and divide the real part and the imaginary part in the signal matrix to determine the signal to be predicted.

[0106] Optionally, the determination module 203 is further used to: input the signal to be predicted into the feature extraction module in the region selection network to perform feature extraction to obtain a feature extraction result; input the feature extraction result into the regression module in the region selection network to determine the target sub-region.

[0107] Optionally, the prediction module 204 is also used to: determine a target position prediction model from multiple position prediction models based on the target sub-area; input the signal to be predicted into the spatial attention module in the target position prediction model, and use the spatial attention module to perform weighted processing on the signal to be predicted to obtain weighted features; input the weighted features into the residual convolutional neural network in the target position prediction model for prediction to obtain position data.

[0108] Optionally, the spatial attention module includes a pooling layer, a convolution layer and an activation function, and the prediction module 204 is also used to: use the pooling layer to perform pooling processing on the signal to be predicted to obtain a pooling processing result; use the convolution layer to perform a convolution operation on the pooling processing result to obtain a convolution processing result; use the activation function to weight the convolution processing result based on a preset weight range to obtain the spatial attention weight; multiply the spatial attention weight by the signal to be predicted element-wise to obtain a weighted feature.

[0109] Optionally, the target position prediction model includes a first convolution module, a second convolution module and a regression module, and the prediction module 204 is also used to: use the first convolution module to perform a first processing on the weighted features to obtain a first processing result, wherein the first processing includes convolution processing, normalization processing, activation processing and pooling processing in sequence; use the second convolution module to perform a second processing on the first processing result to obtain a second processing result, wherein the second processing includes a main branch operation and an auxiliary branch operation; use the regression module to predict the second processing result to obtain position data.

[0110] Optionally, the prediction module 204 is also used to: use the second convolution module to perform a main branch operation on the first processing result to obtain a main branch operation result, wherein the main branch operation includes in sequence: first convolution processing, normalization processing, activation processing, second convolution processing, and second normalization processing; use the second convolution module to perform an auxiliary branch operation on the first processing result to obtain an auxiliary branch operation result, wherein the auxiliary branch operation includes: convolution processing based on a preset dimension; determine the second processing result based on the main branch operation result and the auxiliary branch operation result.

[0111] An embodiment of the present invention further provides an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the positioning method described in any of the above embodiments is executed when the program is running.

[0112] Optionally, in this embodiment, the processor in the electronic device may be configured to run an executable program to perform the following steps:

[0113] Step S101: acquiring a received signal received by an access point, wherein a sensing area of ​​the access point is divided into a plurality of sub-areas.

[0114] Step S102: preprocess the received signal using a preset signal model to obtain a signal to be predicted.

[0115] Step S103, determining a target sub-region corresponding to the signal to be predicted from the perception region according to the signal to be predicted and the pre-trained region selection network.

[0116] Step S104, determining the target position prediction model corresponding to the target sub-region, and inputting the signal to be predicted into the target position prediction model to predict the position data, wherein multiple sub-regions correspond one-to-one to multiple position prediction models.

[0117] Optionally, the processor in the above-mentioned electronic device can be configured to run an executable program to perform the following steps: determine a communication signal expression based on user equipment configuration information and preset access point configuration information of a receiving access point, wherein the user equipment is included in a non-cellular multiple-input multiple-output network; determine a perception signal expression based on preset access point configuration information of a sending access point and a receiving access point, wherein the sending access point and the receiving access point are included in the non-cellular multiple-input multiple-output network; determine a preset signal model based on the communication signal expression, the perception signal expression and additive noise.

[0118] Optionally, the processor in the above-mentioned electronic device can be configured to run an executable program to perform the following steps: determine a communication signal expression based on a communication channel representation and a communication signal representation of the user device and the receiving access point, wherein the communication channel representation and the communication signal representation are included in the user device configuration information and the preset access point configuration information.

[0119] Optionally, the processor in the above-mentioned electronic device can be configured to run an executable program to perform the following steps: determine a perception signal expression based on the perception channel representation and the perception signal representation of the sending access point and the receiving access point, wherein the perception channel representation and the perception signal representation are included in the preset access point configuration information.

[0120] Optionally, the perceived channel representation determined by the processor in the above-mentioned electronic device running the executable program includes: signal attenuation of the transmitting access point and the receiving access point, the dual-base distance between the transmitting access point and the receiving access point, the steering vectors of the transmitting access point and the receiving access point, the subcarrier spacing, the speed of light, the coherent processing interval, the Doppler frequency shift, and the carrier wavelength.

[0121] Optionally, the processor in the above-mentioned electronic device can be configured to run an executable program to perform the following steps: based on a preset signal model, determine multiple perception signals from multiple received signals; determine a signal matrix based on the multiple perception signals, wherein the signal matrix is ​​a complex matrix; divide the real part and the imaginary part in the signal matrix to determine the signal to be predicted.

[0122] Optionally, the processor in the above-mentioned electronic device can be configured to run an executable program to perform the following steps: input the signal to be predicted into the feature extraction module in the region selection network for feature extraction to obtain a feature extraction result; input the feature extraction result into the regression module in the region selection network to determine the target sub-region.

[0123] Optionally, the processor in the above-mentioned electronic device can be configured to run an executable program to perform the following steps: determine a target position prediction model from multiple position prediction models based on the target sub-area; input the signal to be predicted into the spatial attention module in the target position prediction model, and use the spatial attention module to perform weighted processing on the signal to be predicted to obtain weighted features; input the weighted features into the residual convolutional neural network in the target position prediction model for prediction to obtain position data.

[0124] Optionally, the spatial attention module includes a pooling layer, a convolution layer and an activation function, and the processor in the above electronic device can be configured to run an executable program to perform the following steps: using the pooling layer, pooling the signal to be predicted to obtain a pooling processing result; using the convolution layer, convolution operation is performed on the pooling processing result to obtain a convolution processing result; using the activation function, weight restriction is performed on the convolution processing result based on a preset weight range to obtain the spatial attention weight; element-wise multiplication of the spatial attention weight and the signal to be predicted to obtain a weighted feature.

[0125] Optionally, the target position prediction model includes a first convolution module, a second convolution module and a regression module, and the processor in the above electronic device can be configured to run an executable program to perform the following steps: using the first convolution module, performing a first processing on the weighted features to obtain a first processing result, wherein the first processing includes convolution processing, normalization processing, activation processing and pooling processing in sequence; using the second convolution module to perform a second processing on the first processing result to obtain a second processing result, wherein the second processing includes a main branch operation and an auxiliary branch operation; using the regression module, predicting the second processing result to obtain position data.

[0126] Optionally, the processor in the above-mentioned electronic device can be configured to run an executable program to perform the following steps: using the second convolution module to perform a main branch operation on the first processing result to obtain a main branch operation result, wherein the main branch operation includes, in sequence: first convolution processing, normalization processing, activation processing, second convolution processing, and second normalization processing; using the second convolution module to perform an auxiliary branch operation on the first processing result to obtain an auxiliary branch operation result, wherein the auxiliary branch operation includes: convolution processing based on a preset dimension; determining the second processing result based on the main branch operation result and the auxiliary branch operation result.

[0127] An embodiment of the present invention further provides a computer program product, including a computer program, and when the computer program is executed by a processor, the positioning method described in any of the above embodiments is implemented.

[0128] Optionally, in this embodiment, the above computer program implements the following steps when executed by a processor:

[0129] Step S101: acquiring a received signal received by an access point, wherein a sensing area of ​​the access point is divided into a plurality of sub-areas.

[0130] Step S102: preprocess the received signal using a preset signal model to obtain a signal to be predicted.

[0131] Step S103, determining a target sub-region corresponding to the signal to be predicted from the perception region according to the signal to be predicted and the pre-trained region selection network.

[0132] Step S104, determining the target position prediction model corresponding to the target sub-region, and inputting the signal to be predicted into the target position prediction model to predict the position data, wherein multiple sub-regions correspond one-to-one to multiple position prediction models.

[0133] Optionally, the above-mentioned computer program implements the following steps when executed by the processor: determining a communication signal expression based on user equipment configuration information and preset access point configuration information of a receiving access point, wherein the user equipment is included in a non-cellular multiple-input multiple-output network; determining a perception signal expression based on preset access point configuration information of a sending access point and a receiving access point, wherein the sending access point and the receiving access point are included in the non-cellular multiple-input multiple-output network; determining a preset signal model based on the communication signal expression, the perception signal expression and additive noise.

[0134] Optionally, the above-mentioned computer program implements the following steps when executed by the processor: determine the communication signal expression based on the communication channel representation and communication signal representation of the user equipment and the receiving access point, wherein the communication channel representation and the communication signal representation are included in the user equipment configuration information and the preset access point configuration information.

[0135] Optionally, when the computer program is executed by a processor, the following steps are implemented: determining a perception signal expression based on the perception channel representation and the perception signal representation of the sending access point and the receiving access point, wherein the perception channel representation and the perception signal representation are included in the preset access point configuration information.

[0136] Optionally, the perceived channel representation determined by the above-mentioned computer program when executed by the processor includes: signal attenuation of the transmitting access point and the receiving access point, the dual-base distance between the transmitting access point and the receiving access point, the steering vectors of the transmitting access point and the receiving access point, the subcarrier spacing, the speed of light, the coherent processing interval, the Doppler frequency shift, and the carrier wavelength.

[0137] Optionally, the above-mentioned computer program implements the following steps when executed by the processor: based on a preset signal model, determining multiple perception signals from multiple received signals; determining a signal matrix based on the multiple perception signals, wherein the signal matrix is ​​a complex matrix; dividing the real part and the imaginary part in the signal matrix to determine the signal to be predicted.

[0138] Optionally, the above-mentioned computer program implements the following steps when executed by the processor: inputting the signal to be predicted into the feature extraction module in the region selection network for feature extraction to obtain a feature extraction result; inputting the feature extraction result into the regression module in the region selection network to determine the target sub-region.

[0139] Optionally, the above-mentioned computer program implements the following steps when executed by the processor: determine the target position prediction model from multiple position prediction models according to the target sub-area; input the signal to be predicted into the spatial attention module in the target position prediction model, and use the spatial attention module to perform weighted processing on the signal to be predicted to obtain weighted features; input the weighted features into the residual convolutional neural network in the target position prediction model for prediction to obtain position data.

[0140] Optionally, the spatial attention module includes a pooling layer, a convolution layer and an activation function. The above computer program implements the following steps when executed by the processor: using the pooling layer, pooling processing is performed on the signal to be predicted to obtain a pooling processing result; using the convolution layer, convolution operation is performed on the pooling processing result to obtain a convolution processing result; using the activation function, weight restriction is performed on the convolution processing result based on a preset weight range to obtain the spatial attention weight; the spatial attention weight is element-wise multiplied with the signal to be predicted to obtain a weighted feature.

[0141] Optionally, the target position prediction model includes a first convolution module, a second convolution module and a regression module. The above computer program implements the following steps when executed by the processor: using the first convolution module, performing a first processing on the weighted features to obtain a first processing result, wherein the first processing includes convolution processing, normalization processing, activation processing and pooling processing in sequence; using the second convolution module to perform a second processing on the first processing result to obtain a second processing result, wherein the second processing includes a main branch operation and an auxiliary branch operation; using the regression module, predicting the second processing result to obtain position data.

[0142] Optionally, the above-mentioned computer program implements the following steps when executed by the processor: using the second convolution module to perform a main branch operation on the first processing result to obtain a main branch operation result, wherein the main branch operation includes in sequence: first convolution processing, normalization processing, activation processing, second convolution processing, and second normalization processing; using the second convolution module to perform an auxiliary branch operation on the first processing result to obtain an auxiliary branch operation result, wherein the auxiliary branch operation includes: convolution processing based on a preset dimension; determining the second processing result according to the main branch operation result and the auxiliary branch operation result.

[0143] An embodiment of the present invention further provides a computer-readable storage medium, which includes a stored executable program, wherein when the executable program runs, the device where the computer-readable storage medium is located is controlled to execute the positioning method described in any of the above embodiments.

[0144] Optionally, in this embodiment, the above executable program may be configured to store an executable program for performing the following steps:

[0145] Step S101: acquiring a received signal received by an access point, wherein a sensing area of ​​the access point is divided into a plurality of sub-areas.

[0146] Step S102: preprocess the received signal using a preset signal model to obtain a signal to be predicted.

[0147] Step S103, determining a target sub-region corresponding to the signal to be predicted from the perception region according to the signal to be predicted and the pre-trained region selection network.

[0148] Step S104, determining the target position prediction model corresponding to the target sub-region, and inputting the signal to be predicted into the target position prediction model to predict the position data, wherein multiple sub-regions correspond one-to-one to multiple position prediction models.

[0149] Optionally, the above-mentioned executable program can be configured to store an executable program for performing the following steps: determining a communication signal expression based on user equipment configuration information and preset access point configuration information of a receiving access point, wherein the user equipment is included in a non-cellular multiple-input multiple-output network; determining a perception signal expression based on preset access point configuration information of a sending access point and a receiving access point, wherein the sending access point and the receiving access point are included in the non-cellular multiple-input multiple-output network; determining a preset signal model based on the communication signal expression, the perception signal expression and additive noise.

[0150] Optionally, the above-mentioned executable program can be configured to store an executable program for performing the following steps: determining a communication signal expression based on a communication channel representation and a communication signal representation of the user equipment and the receiving access point, wherein the communication channel representation and the communication signal representation are included in the user equipment configuration information and the preset access point configuration information.

[0151] Optionally, the above-mentioned executable program can be configured to store an executable program for performing the following steps: determining a perception signal expression based on the perception channel representation and the perception signal representation of the sending access point and the receiving access point, wherein the perception channel representation and the perception signal representation are included in the preset access point configuration information.

[0152] Optionally, the perceived channel representation determined when the above executable program is executed includes: signal attenuation of the transmitting access point and the receiving access point, the dual-base distance between the transmitting access point and the receiving access point, the steering vectors of the transmitting access point and the receiving access point, the subcarrier spacing, the speed of light, the coherent processing interval, the Doppler frequency shift, and the carrier wavelength.

[0153] Optionally, the above-mentioned executable program can be configured to store an executable program for executing the following steps: based on a preset signal model, determining multiple perception signals from multiple received signals; determining a signal matrix based on the multiple perception signals, wherein the signal matrix is ​​a complex matrix; dividing the real part and the imaginary part in the signal matrix to determine the signal to be predicted.

[0154] Optionally, the above-mentioned executable program can be configured to store an executable program for executing the following steps: inputting the signal to be predicted into a feature extraction module in the region selection network for feature extraction to obtain a feature extraction result; inputting the feature extraction result into a regression module in the region selection network to determine the target sub-region.

[0155] Optionally, the above-mentioned executable program can be configured to store an executable program for executing the following steps: determining a target position prediction model from multiple position prediction models based on the target sub-area; inputting the signal to be predicted into the spatial attention module in the target position prediction model, and using the spatial attention module to perform weighted processing on the signal to be predicted to obtain weighted features; inputting the weighted features into the residual convolutional neural network in the target position prediction model for prediction to obtain position data.

[0156] Optionally, the spatial attention module includes a pooling layer, a convolution layer and an activation function, and the above-mentioned executable program can be configured to store an executable program for executing the following steps: using the pooling layer, pooling the signal to be predicted to obtain a pooling processing result; using the convolution layer, convolution operation on the pooling processing result to obtain a convolution processing result; using the activation function, weight restriction of the convolution processing result based on a preset weight range to obtain the spatial attention weight; element-wise multiplication of the spatial attention weight and the signal to be predicted to obtain a weighted feature.

[0157] Optionally, the target position prediction model includes a first convolution module, a second convolution module and a regression module, and the above-mentioned executable program can be configured to store an executable program for executing the following steps: using the first convolution module, performing a first processing on the weighted features to obtain a first processing result, wherein the first processing includes convolution processing, normalization processing, activation processing and pooling processing in sequence; using the second convolution module to perform a second processing on the first processing result to obtain a second processing result, wherein the second processing includes a main branch operation and an auxiliary branch operation; using the regression module, predicting the second processing result to obtain position data.

[0158] Optionally, the above-mentioned executable program can be configured to store an executable program for executing the following steps: using the second convolution module to perform a main branch operation on the first processing result to obtain a main branch operation result, wherein the main branch operation includes, in sequence: first convolution processing, normalization processing, activation processing, second convolution processing, and second normalization processing; using the second convolution module to perform an auxiliary branch operation on the first processing result to obtain an auxiliary branch operation result, wherein the auxiliary branch operation includes: convolution processing based on a preset dimension; determining the second processing result based on the main branch operation result and the auxiliary branch operation result.

[0159] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0160] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0161] In some embodiments provided in the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0162] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0163] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0164] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0165] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A positioning method, characterized in that: include: Acquire a received signal received by an access point, wherein a sensing area of ​​the access point is divided into a plurality of sub-areas; Preprocessing the received signal using a preset signal model to obtain a signal to be predicted; Determining a target sub-region corresponding to the signal to be predicted from the perception region according to the signal to be predicted and a pre-trained region selection network; A target position prediction model corresponding to the target sub-region is determined, and the signal to be predicted is input into the target position prediction model to predict the position data, wherein the multiple sub-regions correspond one-to-one to multiple position prediction models.

2. The positioning method according to claim 1, characterized in that: The construction of the preset signal model includes: Determining a communication signal expression based on user equipment configuration information and preset access point configuration information of a receiving access point, wherein the user equipment is included in a non-cellular multiple-input multiple-output network; Determine a perception signal expression based on the preset access point configuration information of the sending access point and the receiving access point, wherein the sending access point and the receiving access point are included in a non-cellular multiple-input multiple-output network; The preset signal model is determined according to the communication signal expression, the perception signal expression and additive noise.

3. The positioning method according to claim 2, characterized in that: The determining of the communication signal expression based on the user equipment configuration information and the preset access point configuration information of the receiving access point includes: The communication signal expression is determined based on the communication channel representation and the communication signal representation of the user equipment and the receiving access point, wherein the communication channel representation and the communication signal representation are included in the user equipment configuration information and the preset access point configuration information.

4. The positioning method according to claim 2, characterized in that: The determining the perception signal expression based on preset configuration information of the sending access point and the receiving access point includes: The perception signal expression is determined based on the perception channel representation and the perception signal representation of the sending access point and the receiving access point, wherein the perception channel representation and the perception signal representation are included in the preset access point configuration information.

5. The positioning method according to claim 4, characterized in that: The perceived channel representation includes: signal attenuation of the transmitting access point and the receiving access point, a bistatic distance between the transmitting access point and the receiving access point, steering vectors of the transmitting access point and the receiving access point, subcarrier spacing, speed of light, coherent processing spacing, Doppler shift, and carrier wavelength.

6. The positioning method according to claim 1, characterized in that: The method of preprocessing the received signal by using a preset signal model to obtain a signal to be predicted includes: Based on the preset signal model, determining a plurality of perception signals from the plurality of received signals; Determine a signal matrix according to the multiple perception signals, wherein the signal matrix is ​​a complex matrix; The real part and the imaginary part in the signal matrix are divided to determine the signal to be predicted.

7. The positioning method according to claim 1, characterized in that: The step of determining a target sub-region corresponding to the signal to be predicted from the perception region according to the signal to be predicted and a pre-trained region selection network includes: Inputting the signal to be predicted into the feature extraction module in the region selection network to perform feature extraction, thereby obtaining a feature extraction result; The feature extraction result is input into the regression module in the region selection network to determine the target sub-region.

8. The positioning method according to claim 1, characterized in that: The step of determining a target position prediction model corresponding to the target sub-region and inputting the signal to be predicted into the target position prediction model to predict and obtain position data includes: Determining the target position prediction model from the multiple position prediction models according to the target sub-area; Inputting the signal to be predicted into the spatial attention module in the target position prediction model, and using the spatial attention module to perform weighted processing on the signal to be predicted to obtain a weighted feature; The weighted features are input into the residual convolutional neural network in the target position prediction model for prediction to obtain the position data.

9. The positioning method according to claim 8, characterized in that: The spatial attention module includes a pooling layer, a convolution layer and an activation function. The signal to be predicted is input into the spatial attention module in the target position prediction model, and the signal to be predicted is weighted by the spatial attention module to obtain a weighted feature, including: Using the pooling layer, performing pooling processing on the signal to be predicted to obtain a pooling processing result; Using the convolution layer, performing a convolution operation on the pooling processing result to obtain a convolution processing result; Using the activation function, weight limiting the convolution processing result based on a preset weight range to obtain a spatial attention weight; The spatial attention weight is element-wise multiplied by the signal to be predicted to obtain the weighted feature.

10. The positioning method according to claim 8, characterized in that: The target position prediction model includes a first convolution module, a second convolution module and a regression module, and the weighted features are input into the residual convolution neural network in the target position prediction model for prediction to obtain the position data, including: Using the first convolution module, performing a first processing on the weighted features to obtain a first processing result, wherein the first processing sequentially includes convolution processing, normalization processing, activation processing, and pooling processing; Using the second convolution module to perform a second processing on the first processing result to obtain a second processing result, wherein the second processing includes a main branch operation and an auxiliary branch operation; The second processing result is predicted by using the regression module to obtain the position data.

11. The positioning method according to claim 10, characterized in that: The second convolution module is used to perform a second process on the first processing result to obtain a second processing result, including: Using the second convolution module to perform a main branch operation on the first processing result to obtain a main branch operation result, wherein the main branch operation sequentially includes: a first convolution process, a normalization process, an activation process, a second convolution process, and a second normalization process; Using the second convolution module to perform an auxiliary branch operation on the first processing result to obtain an auxiliary branch operation result, wherein the auxiliary branch operation includes: convolution processing based on a preset dimension; The second processing result is determined according to the main branch operation result and the auxiliary branch operation result.

12. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 11 when running.

13. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the storage medium is located is controlled to execute the method according to any one of claims 1 to 11.

Images