AOA and TOF joint estimation method, device and storage medium for indoor positioning

Abstract

The present invention discloses a method, device, and storage medium for joint estimation of AOA and TOF for indoor positioning. The method comprises preprocessing acquired CSI data to obtain processed data; inputting the processed data into a pretrained joint estimation model to obtain the AOA and TOF corresponding to the processed data; and training the joint estimation model using a deep convolutional neural network that utilizes convolution kernels of different sizes in parallel. The present invention utilizes a deep convolutional neural network constructed using convolution kernels of different sizes in parallel to train the joint estimation model, thereby improving the accuracy, resolution, and noise immunity of the estimation results.

Description

Technical Field

[0001] The present invention relates to an AOA and TOF joint estimation method, device and storage medium for indoor positioning, and belongs to the technical field of wireless signal processing. Background Art

[0002] Wireless signals are widely used in various areas of daily life. In areas such as health awareness, fire location rescue, and augmented reality-based navigation, accurately estimating the direction of arrival (AOA) and time of arrival (TOF) of each wireless signal path is crucial. However, during multipath propagation of wireless signals, objects near indoor APs and mobile clients can reflect wireless signals, resulting in low AOA and TOF estimation accuracy.

[0003] Currently, most studies have constructed mathematical models and combined them with signal processing algorithms to obtain a correspondence between received signals and AOA. However, a large number of approximations are made during the model establishment and solution process, and the approximate operations will lose some information, which has a certain impact on the accurate estimation of AOA. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method, device and storage medium for joint estimation of AOA and TOF for indoor positioning.

[0005] In a first aspect, the present invention provides a method for joint estimation of AOA and TOF for indoor positioning, the method comprising:

[0006] Preprocessing the acquired CSI data to obtain processed data;

[0007] The processed data is input into the pre-trained joint estimation model to obtain the AOA and TOF corresponding to the processed data;

[0008] The joint estimation model is obtained by training a deep convolutional neural network, and the deep convolutional neural network uses convolution kernels of different sizes in parallel.

[0009] Furthermore, the acquired CSI data is represented as a complex data matrix image, and the processed data is represented as a real number matrix image.

[0010] Furthermore, the processed data is represented as a three-channel real matrix image.

[0011] Furthermore, the acquisition of the three-channel real matrix image includes:

[0012] The complex data matrix of the acquired CSI data and its conjugate matrix are divided into four sub-matrices according to the real part and the imaginary part and reorganized to obtain a reorganized matrix;

[0013] The reorganized matrix is ​​shifted and intercepted by sliding windows with sliding steps of 2, 4 and 6, respectively, to obtain several sub-matrices, and then the several sub-matrices obtained are spliced ​​to obtain three matrices. The three matrices constitute the obtained three-channel real matrix image.

[0014] Furthermore, the deep convolutional neural network includes an input layer, a 7*7 convolutional layer, a maximum pooling layer, a 1*1 convolutional layer, a 3*3 convolutional layer, a pooling layer, 2 Inception structures, a maximum pooling layer, 3 Inception structures, a maximum pooling layer, 2 Inception structures, an average pooling layer, a fully connected layer, and an output layer.

[0015] Furthermore, the Inception structure includes four layers: the first layer is the input layer, the second layer is a 1*1 convolution layer, a 1*1 convolution layer, a 3*3 maximum pooling layer and a 1*1 convolution layer, the third layer is a 3*3 convolution layer, a 5*5 convolution layer and a 1*1 convolution layer, and the fourth layer is the output layer.

[0016] Furthermore, the training of the joint estimation model includes:

[0017] Obtain historical CSI data and the corresponding AOA and TOF to construct a data set;

[0018] Preprocessing the data set to construct a training set;

[0019] The training set is trained through a deep convolutional neural network to obtain a joint estimation model.

[0020] In a second aspect, the present invention provides an apparatus comprising a processor and a storage medium;

[0021] The storage medium is used to store instructions;

[0022] The processor is configured to operate according to the instructions to execute the steps of the method of the first aspect.

[0023] In a third aspect, the present invention provides a storage medium having a computer program stored thereon, which implements the steps of the method described in the first aspect when executed by a processor.

[0024] Compared with the existing technology, the beneficial effects of the present invention are: the present invention obtains a joint estimation model by training a deep convolutional neural network constructed in parallel based on convolution kernels of different sizes, thereby achieving higher accuracy, higher resolution and better noise resistance than the results estimated by traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1This is a flowchart of the AOA and TOF joint estimation method for indoor positioning according to an embodiment of the present invention;

[0026] Figure 2 is a schematic diagram of an antenna array at a receiving end according to an embodiment of the present invention;

[0027] Figure 3 2 is a schematic diagram of a single-channel construction method for CSI data in a data set according to an embodiment of the present invention;

[0028] Figure 4 Schematic diagram of a multi-channel construction method for CSI data in a data set according to an embodiment of the present invention;

[0029] Figure 5 2 is a schematic diagram of a method for convolving a multi-scale convolution kernel with CSI data according to an embodiment of the present invention;

[0030] Figure 6 This is a diagram of a neural network structure designed with convolution kernels of different sizes connected in parallel according to an embodiment of the present invention;

[0031] Figure 7 This is a comparison chart of the root mean square error (RMSE) of the combined estimation value of the embodiment of the present invention and the estimation results of various methods;

[0032] Figure 8 This is a comparison diagram of the error distribution of the joint estimation value of the embodiment of the present invention and the estimation results of various methods. DETAILED DESCRIPTION

[0033] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0034] In the description of the present invention, "several" means more than one, "plurality" means more than two, "greater than," "less than," and "exceed" are understood to exclude the number itself, while "above," "below," and "within" are understood to include the number itself. The use of "first" and "second" in the description is solely for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, implicitly specifying the number of the indicated technical features, or implicitly specifying the order of the indicated technical features.

[0035] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installing, and connecting should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention based on the specific content of the technical solution.

[0036] In the description of the present invention, reference to terms such as "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the exemplary expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0037] The present invention provides a method, device, and storage medium for joint estimation of AOA and TOF for indoor positioning. The present invention is further described below with reference to the accompanying drawings and embodiments, wherein:

[0038] Example 1:

[0039] like Figure 1 As shown, an embodiment of the present invention provides an AOA and TOF joint estimation method for indoor positioning, including:

[0040] Use OFDM signal as the transmission signal, set the transmission channel to Gaussian channel, the channel bandwidth is 40MHz, the center frequency is 5.32GHz, the adjacent subcarrier spacing is set to 312.5KHz, and 30 subcarriers are selected, such as Figure 2 As shown, the receiving end uses a linear array with three antennas and generates a data set by setting different channel parameters. The data set samples include input data and labels. The received CSI is used as input data, and AOA and TOF are used as labels.

[0041] The CSI data in the dataset is a complex matrix. However, the current deep convolutional neural network does not have a more appropriate way to process the complex input. In order to obtain a CSI image that can be used as input, the present invention converts the CSI complex data matrix H received from the antenna array into 90×1 Divided into H R , H I , where H R is the real part matrix of H, H I is the imaginary matrix of H, and then transform it by formula (1) to obtain It can effectively retain information such as the amplitude and phase of CSI;

[0042]

[0043] Among them csi m，k，m＝1，2，3， k＝1，2，...，30 Re(csi) represents the kth subcarrier on the mth antenna, which is complex data. Re(csi) represents the real part of csi, and Im(csi) represents the imaginary part of csi.

[0044] The convolutional neural network extracts the features of the input image through the convolution kernel, considering that the estimated values ​​of TOF and AOA are related to multiple subcarriers, such as Figure 3 As shown, the present invention proposes a method for constructing a single-channel input matrix, let h i，j Representation matrix The value of the jth (j=1,2) column in the i-th (i=1,2,…,180) row is set to 2 so that the convolution kernel can cover multiple subcarriers when training the network. When the window moves, the front data is used to fill the tail end. It represents the matrix obtained by moving the window i times. Finally, these 90 matrices with a size of 180×2 are merged into a new matrix:

[0045]

[0046] In order to extract more features of each subcarrier signal combination, such as Figure 4 As shown, the present invention proposes a method for constructing a multi-channel input matrix, changing the moving step size to 4 and 6 respectively, and then performing the same filling operation on the window to obtain two two-dimensional matrices The three matrices obtained are combined into a three-dimensional matrix H as the input of the neural network 180×180×3 .

[0047] A training set is constructed based on the three-channel real matrix image obtained through processing and the AOA and TOF corresponding to the three-channel real matrix image.

[0048] The deep convolutional neural network is designed based on the parallel connection of convolution kernels of different sizes. Convolution kernels of different scales can be used according to Figure 5 The method represented performs convolution operation with the received signals of different frequency subcarriers, and then extracts the data features related to each subcarrier, such as Figure 6 As shown, the deep convolutional neural network includes an input layer, a 7*7 convolutional layer, a maximum pooling layer, a 1*1 convolutional layer, a 3*3 convolutional layer, a pooling layer, 2 Inception structures, a maximum pooling layer, 3 Inception structures, a maximum pooling layer, 2 Inception structures, an average pooling layer, a fully connected layer, and an output layer; the Inception structure includes four layers: the first layer is the input layer, the second layer is a 1*1 convolutional layer, a 1*1 convolutional layer, a 3*3 maximum pooling layer and a 1*1 convolutional layer, the third layer is a 3*3 convolutional layer, a 5*5 convolutional layer and a 1*1 convolutional layer, and the fourth layer is the output layer.

[0049] The training set is trained through a deep convolutional neural network to obtain a joint estimation model.

[0050] The CSI data to be estimated is preprocessed to obtain a three-channel real matrix image. The three-channel real matrix is ​​used as the input of the joint estimation model, and the AOA and TOF corresponding to the CSI data to be estimated are output.

[0051] like Figure 7 As shown in the figure, it is a comparison diagram of the root mean square error (RMSE) of the estimation results of the method of the present invention and other algorithms. It can be seen that compared with the method proposed in the present invention, the RMSE of the AOA and TOF estimation of the other two methods (SpoFi and Join-2D) decreases with the increase of SNR; since the training samples include data under multiple signal-to-noise ratios, the trained network has good generalization ability, so it has more stable estimation results under different signal-to-noise ratios; from the results, it can be seen that the method proposed in this article performs better than the other two methods at any signal-to-noise ratio.

[0052] like Figure 8 As shown, taking the AOA estimation error distribution in the left figure as an example, from the error distribution point of view, the AOA estimation errors of the present invention are all distributed within 10°, while the data volume with estimation errors within 10° of the other two methods only accounts for 80% and 70% of the total data, and nearly 20% of the data has an estimation error of more than 30°, that is, it cannot be accurately identified.

[0053] Another advantage of the method proposed in the present invention is that even if the arrival angles and arrival times of the two paths are similar, they can still be accurately identified, while the recognition performance of the other two in this case is not ideal. This is because the estimation algorithms in SpoFi and Join-2D are both improved algorithms based on the MUSIC algorithm, in which the MUSIC algorithm uses the orthogonality of the signal subspace and the noise subspace of the signal to construct a spatial spectrum function and estimates the AOA by searching the spectrum peaks. If the spectrum peaks are close, it is difficult to distinguish them, so the traditional algorithm has certain resolution limitations. The neural network-based method proposed in the present invention estimates the AOA and TOF by extracting the relevant features of the signal. The estimated values ​​between the various paths do not affect each other, so there is no problem of insufficient resolution. In the case of Table 1, SpoFi and Join-2D can only identify one path due to their low resolution, while the method proposed in the present invention can accurately identify both paths within a certain error range.

[0054] Table 1 Estimation results under similar AOA paths (null indicates recognition failure)

[0055] True value SpoFi Join-2D Method of the present invention Path 1: (123°, 96ns) (125°, 107ns) (124°, 99ns) (123.6°, 85ns) Path 2: (132°, 102ns) null null (133.7°, 94ns)

[0056] Example 2:

[0057] This embodiment provides a device including a processor and a storage medium;

[0058] The storage medium is used to store instructions;

[0059] The processor is configured to operate according to the instructions to execute the steps of the method described in Example 1.

[0060] Example 3:

[0061] This embodiment provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described in Example 1 are implemented.

[0062] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0063] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0064] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0065] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1A step that specifies a function in one or more boxes.

[0066] 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 technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. AOA and TOF joint estimation method for indoor positioning, characterized by: The method comprises: Preprocessing the acquired CSI data to obtain processed data; Inputting the processed data into a pre-trained joint estimation model to obtain the AOA and TOF corresponding to the processed data; the joint estimation model is obtained by training a deep convolutional neural network, and the deep convolutional neural network uses convolution kernels of different sizes in parallel; The acquired CSI data is represented as a complex data matrix image, and the processed data is represented as a real matrix image; the processed data is represented as a three-channel real matrix image; The acquisition of the three-channel real matrix image includes: The complex data matrix of the acquired CSI data and its conjugate matrix are divided into four sub-matrices according to the real part and the imaginary part and reorganized to obtain a reorganized matrix; The reorganized matrix is ​​shifted and intercepted by sliding windows with sliding steps of 2, 4 and 6, respectively, to obtain several sub-matrices, and then the several sub-matrices obtained are spliced ​​to obtain three matrices. The three matrices constitute the obtained three-channel real matrix image.

2. The AOA and TOF joint estimation method for indoor positioning according to claim 1, characterized in that: The deep convolutional neural network includes an input layer, a 7*7 convolutional layer, a maximum pooling layer, a 1*1 convolutional layer, a 3*3 convolutional layer, a pooling layer, two Inception structures, a maximum pooling layer, three Inception structures, a maximum pooling layer, two Inception structures, an average pooling layer, a fully connected layer, and an output layer.

3. The AOA and TOF joint estimation method for indoor positioning according to claim 2, characterized in that: The Inception structure consists of four layers: the first layer is the input layer, the second layer is a 1*1 convolution layer, a 1*1 convolution layer, a 3*3 maximum pooling layer and a 1*1 convolution layer, the third layer is a 3*3 convolution layer, a 5*5 convolution layer and a 1*1 convolution layer, and the fourth layer is the output layer.

4. The AOA and TOF joint estimation method for indoor positioning according to claim 1, characterized in that: The training of the joint estimation model includes: Obtain historical CSI data and the corresponding AOA and TOF to construct a data set; Preprocessing the data set to construct a training set; The training set is trained through a deep convolutional neural network to obtain a joint estimation model.

5. A device, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 4.

6. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images