PDP-based behavior recognition method and device, electronic device, and storage medium

Abstract

The present application provides a behavior recognition method and device based on PDP, an electronic device, and a storage medium, including: for the channel estimation information of the Wi-Fi signal, converting the channel estimation information into time domain channel estimation information by inverse Fourier transform; determining the covariance matrix of the time domain channel estimation information, and performing eigenvalue decomposition on the covariance matrix to obtain multiple eigenvalues; determining the target number of effective channels based on the multiple eigenvalues; determining the power of multiple effective channels from the covariance matrix based on the target number; removing the power of the LOS channel from the power of multiple effective channels to obtain the power of multiple NLOS channels; performing behavior recognition based on the power of the multiple NLOS channels to obtain a behavior recognition result. The solution of the present application accurately realizes the behavior recognition of passive targets with the help of the power delay spectrum of the Wi-Fi signal.

Description

Technical Field

[0001] The present application relates to the field of behavior recognition technology, and in particular to a behavior recognition method and device based on PDP, an electronic device, and a computer-readable storage medium. Background Art

[0002] In the future, Wi-Fi signals will not only have data transmission functions, but will also have the ability to perceive passive targets. Passive targets refer to people or other targets that do not carry terminal devices. For example, the channel state information (CSI) of the LTF (Long Training Field) of the Wi-Fi signal is used to locate, track, recognize gestures, and detect health behaviors (such as heartbeat, breathing). Using Wi-Fi signals for behavior recognition can play an important role in smart home scenarios, allowing various smart terminals to have better control characteristics, helping users get rid of remote controls and gain a better long-distance control experience. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a behavior recognition method and device based on PDP, an electronic device, and a computer-readable storage medium, which are used to accurately realize the behavior recognition of passive targets by means of the power delay spectrum of Wi-Fi signals.

[0004] On the one hand, the present application provides a behavior recognition method based on PDP, comprising:

[0005] For the channel estimation information of the Wi-Fi signal, convert the channel estimation information into time domain channel estimation information by inverse Fourier transform;

[0006] Determine a covariance matrix of the time domain channel estimation information, and perform eigenvalue decomposition on the covariance matrix to obtain a plurality of eigenvalues;

[0007] Determining a target number of valid channels based on the plurality of characteristic values;

[0008] Based on the target number, determining the power of a plurality of effective channels from the covariance matrix;

[0009] Subtracting the power of the LOS channel from the power of the multiple effective channels to obtain the power of the multiple NLOS channels;

[0010] Based on the power of the multiple NLOS channels, behavior recognition is performed to obtain a behavior recognition result.

[0011] In one embodiment, determining a target number of valid channels based on the multiple characteristic values ​​includes:

[0012] Arrange the multiple eigenvalues ​​in descending order to obtain an eigenvalue sequence; wherein the eigenvalue sequence includes eigenvalues ​​numbered 0 to M; M is a positive integer;

[0013] Splitting the eigenvalue sequence into a first subsequence and a second subsequence; wherein the first subsequence includes the eigenvalues ​​numbered 1 to M-1 in the eigenvalue sequence, and the second subsequence includes the eigenvalues ​​numbered 2 to M in the eigenvalue sequence;

[0014] Subtracting the second subsequence from the first subsequence to obtain a difference sequence; wherein the difference sequence includes the difference values ​​from sequence number 1 to M-1;

[0015] The maximum difference is found from the difference sequence, and the target number of valid channels is determined according to the sequence number of the found maximum difference.

[0016] In one embodiment, determining the power of multiple effective channels from the covariance matrix based on the target number includes:

[0017] Arrange multiple elements on the diagonal of the covariance matrix in order from large to small;

[0018] A target number of elements at the front of the sort are selected as the power of multiple effective channels.

[0019] In one embodiment, before converting the channel estimation information for the Wi-Fi signal into time-domain channel estimation information by inverse Fourier transform, the method further includes:

[0020] By using a boundary repetition method, the number of frequency domain channel estimation values ​​corresponding to the subcarriers in the channel estimation information is supplemented to the number of specified sampling points;

[0021] A circular shift operation is performed on the complemented channel estimation information.

[0022] In one embodiment, the performing behavior recognition based on the power of the plurality of NLOS channels to obtain the behavior recognition result includes:

[0023] A two-dimensional matrix is ​​constructed according to the power of multiple NLOS channels corresponding to multiple consecutive data packets;

[0024] Converting the two-dimensional matrix into an image, and cutting out a sub-image to be processed from the image;

[0025] The sub-image to be processed is input into a trained behavior recognition model to obtain a behavior recognition result output by the behavior recognition model.

[0026] In one embodiment, cutting out a sub-image to be processed from the image includes:

[0027] A sub-image generated within a specified time period is cut out from the image as the sub-image to be processed.

[0028] In one embodiment, before inputting the sub-image to be processed into a trained behavior recognition model to obtain a behavior recognition result output by the behavior recognition model, the method further includes:

[0029] Acquire multiple sample channel estimation information of Wi-Fi signals, and convert the multiple sample channel estimation information into sample images respectively; wherein the multiple sample channel estimation information corresponds to multiple preset behavior categories;

[0030] According to each sample image, multiple sample sub-images are obtained by cutting, and a behavior category label corresponding to the sample image is added to each sample sub-image;

[0031] The machine learning model is trained based on a plurality of sample sub-images carrying behavior category labels to obtain the behavior recognition model.

[0032] On the other hand, the present application provides a behavior recognition device based on PDP, comprising:

[0033] A transformation module, configured to transform the channel estimation information of the Wi-Fi signal into time-domain channel estimation information by inverse Fourier transformation;

[0034] A decomposition module, used to determine the covariance matrix of the time domain channel estimation information, and perform eigenvalue decomposition on the covariance matrix to obtain multiple eigenvalues;

[0035] A first determination module, configured to determine a target number of valid channels based on the multiple characteristic values;

[0036] A second determination module is used to determine the power of multiple effective channels from the covariance matrix based on the target number;

[0037] A removal module, used for removing the power of the LOS channel from the power of the multiple effective channels to obtain the power of the multiple NLOS channels;

[0038] The identification module is used to perform behavior identification based on the power of the multiple NLOS channels to obtain a behavior identification result.

[0039] Furthermore, the present application provides an electronic device, the electronic device comprising:

[0040] processor;

[0041] a memory for storing processor-executable instructions;

[0042] Wherein, the processor is configured to execute the above-mentioned behavior recognition method based on PDP.

[0043] In addition, the present application provides a computer-readable storage medium, which stores a computer program. The computer program can be executed by a processor to complete the above-mentioned PDP-based behavior recognition method.

[0044] The solution of this application can convert the channel state information of the Wi-Fi signal into PDP, and remove the noise and the LOS channel as the background, so as to obtain the power of the NLOS channel that can accurately characterize the activity state of the passive target, thereby achieving accurate behavior recognition. Since PDP is stronger than CSI, it is easier to characterize the impact of passive targets and eliminates the extremely strong LOS background. Therefore, behavior recognition using the PDP of the NLOS channel has good generalization ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solution of the embodiments of the present application, the drawings required for use in the embodiments of the present application are briefly introduced below.

[0046] Figure 1 A schematic diagram of the structure of an electronic device provided in one embodiment of the present application;

[0047] Figure 2 A flow chart of a behavior recognition method based on PDP provided in one embodiment of the present application;

[0048] Figure 3 An embodiment of the present application provides Figure 2 Detailed flow chart of step 230;

[0049] Figure 4 A schematic diagram of a boundary repetition method provided in an embodiment of the present application;

[0050] Figure 5 An embodiment of the present application provides Figure 2 Detailed flow chart of step 260;

[0051] Figure 6 A schematic diagram of a two-dimensional matrix provided in one embodiment of the present application;

[0052] Figure 7 A flowchart of a method for training a behavior recognition model provided in one embodiment of the present application;

[0053] Figure 8 A block diagram of a PDP-based behavior recognition device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0055] Similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0056] like Figure 1 As shown, this embodiment provides an electronic device 1, including: at least one processor 11 and a memory 12, Figure 1 A processor 11 is taken as an example. The processor 11 and the memory 12 are connected via a bus 10. The memory 12 stores instructions that can be executed by the processor 11. The instructions are executed by the processor 11 so that the electronic device 1 can execute all or part of the process of the method in the following embodiment. In one embodiment, the electronic device 1 can be a router, a communication terminal device (such as a mobile phone, a tablet computer), etc., which is used to execute a behavior recognition method based on PDP. The electronic device can be equipped with a transmitting antenna and a receiving antenna for transmitting and receiving Wi-Fi signals. The following describes the scheme with the electronic device as the execution subject.

[0057] The memory 12 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable red-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, disk or optical disk.

[0058] The present application also provides a computer-readable storage medium, which stores a computer program. The computer program can be executed by the processor 11 to complete the PDP-based behavior recognition method provided by the present application.

[0059] In an indoor environment, the electronic device that implements the solution of the present application can be placed in a specified position in the indoor scene (for example, the electronic device is a router, which is placed in the center of the TV wall in the living room). In the process of transmitting and receiving Wi-Fi signals by the electronic device, channel estimation information as channel state information can be obtained through channel estimation. Since there are various objects in the indoor scene (such as people, furniture, electrical appliances, pets, etc.), various reflections will occur in the Wi-Fi signal. Therefore, there are multiple wireless channels between the transmitter and the receiver. When a passive target moves in the scene (such as running, making gestures, falling, etc.), since the indoor static objects are kept motionless, the channel state information can characterize the impact of the behavior of the passive target on the wireless channel. Therefore, the behavior of the passive target in the indoor scene can be identified with the help of the channel state information.

[0060] See also Figure 2 , is a flow chart of a behavior recognition method based on PDP provided in an embodiment of the present application, such as Figure 2 As shown, the method may include the following steps 210 - 260 .

[0061] Step 210: For the channel estimation information of the Wi-Fi signal, convert the channel estimation information into time domain channel estimation information by inverse Fourier transform.

[0062] During the transmission and reception of Wi-Fi signals, the electronic device may obtain channel estimation information of the Wi-Fi signal by means of channel estimation. The channel estimation information estimates the channel state information of the Wi-Fi signal. For example, the electronic device may determine the channel estimation information in the frequency domain from the pilot signal of the Wi-Fi signal by means of the LS (Least Square) algorithm. In one embodiment, in order to make the frequency leakage more concentrated, the channel estimation information may be windowed by means of a window function (e.g., a Hamming window).

[0063] After obtaining the channel estimation information, the electronic device may transform the channel estimation information into the time domain by inverse Fourier transform (IFFT) to obtain the time domain channel estimation information. Exemplarily, the transformation process may refer to the following formula (1):

[0064]

[0065] in, is the channel estimation information in the frequency domain; is the time domain channel estimation information; q is the receiving antenna index; k is the subcarrier index; l is the symbol index.

[0066] Step 220: Determine the covariance matrix of the time domain channel estimation information, and perform eigenvalue decomposition on the covariance matrix to obtain multiple eigenvalues.

[0067] The time domain channel estimation information includes multiple time domain channel estimation values, which can be regarded as a matrix. The electronic device can determine the corresponding covariance matrix based on the time domain information estimation information. Exemplarily, the covariance matrix can be determined by the following formula (2):

[0068] R hh =E[hh ct ] (2)

[0069] Among them, R hh is the covariance matrix; h is the time domain channel estimation information in the form of a column vector.

[0070] After obtaining multiple covariance matrices, eigenvalue decomposition can be performed on the covariance matrices to obtain multiple eigenvalues, where each eigenvalue corresponds to a channel.

[0071] Step 230: Determine a target number of effective channels based on multiple characteristic values.

[0072] After obtaining multiple eigenvalues, the target number of effective channels can be determined according to the size distribution law of the eigenvalues. Here, the effective channel refers to the channel for multipath transmission of Wi-Fi signals.

[0073] In one embodiment, see Figure 3 , provided in one embodiment of the present application Figure 2 The detailed flow chart of step 230 is as follows: Figure 3 As shown, when executing step 230, the following steps 231 to 234 may be specifically executed.

[0074] Step 231: Arrange multiple eigenvalues ​​in descending order to obtain an eigenvalue sequence; wherein the eigenvalue sequence contains eigenvalues ​​numbered 0 to M; and M is a positive integer.

[0075] In this application, the total number of eigenvalues ​​is recorded as M+1. After arranging multiple eigenvalues ​​in order from large to small, a eigenvalue sequence can be obtained, and the sequence number of each eigenvalue in the eigenvalue sequence is determined according to the order of arrangement. The sequence number starts from 0 to M. The larger the sequence number of the eigenvalue, the smaller the eigenvalue.

[0076] Step 232: Split the eigenvalue sequence into a first subsequence and a second subsequence; wherein the first subsequence includes eigenvalues ​​numbered 1 to M-1 in the eigenvalue sequence, and the second subsequence includes eigenvalues ​​numbered 2 to M in the eigenvalue sequence.

[0077] The electronic device can separate the first subsequence and the second subsequence from the eigenvalue sequence. It should be noted that the eigenvalue with sequence number 0 corresponds to the LOS (Line of Sight) channel, and the size of this eigenvalue is quite different from other eigenvalues. In order to avoid the eigenvalue affecting the subsequent processing process, the first subsequence and the second subsequence separated do not contain this eigenvalue.

[0078] For example, from the eigenvalue sequence [a0, a1, ..., a M ] splits the first subsequence [a1, a2, ..., a M-1 ] and the second subsequence [a2, a3, ..., a M ]. Where a represents the eigenvalue.

[0079] Step 233: Subtract the second subsequence from the first subsequence to obtain a differential sequence; wherein the differential sequence includes the difference values ​​from sequence number 1 to M-1.

[0080] The electronic device can subtract each characteristic value in the first subsequence from each characteristic value in the second subsequence one by one, and form a differential sequence according to the multiple difference values. For example, the first subsequence [a1, a2, ..., a M-1 ] and the second subsequence [a2, a3, ..., a M ], when subtracting, a1 minus a2, a2 ​​minus a3, and so on, a M-1 minus a M , the difference sequence can be expressed as [b1, b2, ..., b M-1 ].

[0081] Step 234: Find the maximum difference from the differential sequence, and determine the target number of valid channels based on the sequence number of the maximum difference found.

[0082] Each difference in the differential sequence is obtained by subtracting the next eigenvalue from the previous eigenvalue in the eigenvalue sequence. For the effective channel, the corresponding eigenvalue is larger, while the eigenvalue corresponding to the noise channel is smaller. Therefore, the maximum difference in the differential sequence is obtained by subtracting the eigenvalue corresponding to the first noise channel from the eigenvalue of the last effective channel. Since the eigenvalue with sequence number 0 has been excluded before, it can be known that the sequence number of the maximum difference at this time is the total number of NLOS (Non Line of Sight) channels.

[0083] After finding the maximum difference, the sequence number of the maximum difference is increased by 1, and the target number of effective channels can be obtained. Here, the effective channels include LOS channels and NLOS channels.

[0084] Step 240: Based on the target quantity, determine the power of multiple effective channels from the covariance matrix.

[0085] The power of the effective channel can be obtained by the square of the time domain channel estimation value, and according to the calculation method of the covariance matrix, it can be known that each element on the diagonal of the covariance matrix is ​​the power of each channel. In this case, the electronic device can sort the multiple elements on the diagonal of the covariance matrix from large to small, and select the target number of elements at the front after sorting as the power of multiple effective channels.

[0086] Since the power of the effective channel is much greater than the power of the noise channel, after sorting multiple elements on the diagonal of the covariance matrix, the power of the effective channel can be selected from the multiple elements based on the total amount of the effective channels.

[0087] Step 250: Subtract the power of the LOS channel from the power of the multiple effective channels to obtain the power of the multiple NLOS channels.

[0088] Since the passive target has no effect on the LOS channel when it is active, the power of the LOS channel can be removed so as to focus on the NLOS channel affected by the passive target. The electronic device can remove the maximum power from the power of multiple effective channels, thereby removing the power of the LOS channel and obtaining the power of the remaining NLOS channel.

[0089] Step 260: Based on the power of multiple NLOS channels, behavior recognition is performed to obtain a behavior recognition result.

[0090] For each symbol in each data packet transmitted by the Wi-Fi signal, the power of multiple NLOS channels can be obtained through the above-mentioned process. After calculating the average of the power of each symbol in a data packet in each NLOS channel, the power of multiple NLOS channels corresponding to the data packet can be obtained. When each data packet is continuously processed, the power of multiple NLOS channels corresponding to each data packet is obtained. These powers can be regarded as the PDP (Power Delay Profile) sampling point values ​​of the Wi-Fi signal. Compared with the CSI of the Wi-Fi signal, the PDP has a greater intensity and a higher degree of discreteness, and can more significantly characterize the impact of passive target activities. Therefore, accurate behavior recognition results can be obtained by using the power of multiple NLOS channels for behavior recognition.

[0091] Through the above measures, the channel state information of the Wi-Fi signal can be converted into PDP, and the noise and LOS channel as the background can be removed, so as to obtain the power of the NLOS channel that can accurately characterize the activity state of the passive target, thereby achieving accurate behavior recognition. Since PDP is stronger than CSI, it is easier to characterize the impact of passive targets and eliminates the extremely strong LOS background. Therefore, behavior recognition using the PDP of the NLOS channel has good generalization ability.

[0092] In one embodiment, before executing step 210, the channel estimation information may be processed to obtain a better behavior recognition effect. The electronic device may use a boundary repetition method to supplement the number of frequency domain channel estimation values ​​corresponding to the subcarriers in the channel estimation information to the specified number of sampling points. Here, the specified number of sampling points may be preconfigured as needed.

[0093] See also Figure 4 , is a schematic diagram of a boundary repetition method provided in an embodiment of the present application, such as Figure 4 As shown, the channel estimation information includes frequency domain channel estimation values ​​of 60 subcarriers, the frequency domain channel estimation value of the first subcarrier is repeated 34 times, and the frequency domain channel estimation value of the last subcarrier is repeated 34 times, so that the subcarriers in the channel estimation information are supplemented to 128. Figure 4 The numerical value in is an example. In actual applications, the number of subcarriers and the number of specified sampling points in the channel estimation information are not limited in this application.

[0094] After the channel estimation information is supplemented with the number of sampling points, a circular shift operation may be performed on the supplemented channel estimation information.

[0095] Through the above measures, the channel estimation information can be de-noised and the phase deviation can be eliminated.

[0096] In one embodiment, see Figure 5 , provided in one embodiment of the present application Figure 2 The detailed flow chart of step 260 is as follows: Figure 5 As shown, when executing step 260, the following steps 261 to 263 may be specifically executed.

[0097] Step 261: construct a two-dimensional matrix according to the powers of multiple NLOS channels corresponding to multiple consecutive data packets.

[0098] The electronic device can use the power of multiple NLOS channels corresponding to a data packet as a column in a two-dimensional matrix, and use the power of multiple NLOS channels of multiple consecutive data packets as multiple columns of the two-dimensional matrix, thereby forming a two-dimensional matrix.

[0099] See also Figure 6 , is a schematic diagram of a two-dimensional matrix provided in an embodiment of the present application, such as Figure 6 As shown, each data packet corresponds to the PDP sampling points (power) of L-1 NLOS channels. As the observation time goes by, the electronic device can continuously update the columns of the two-dimensional matrix.

[0100] Step 262: Convert the two-dimensional matrix into an image, and cut out a sub-image to be processed from the image.

[0101] After obtaining the two-dimensional matrix, the electronic device may convert the two-dimensional matrix into an image, and the image format may be but is not limited to HSV (Hue, Saturation, Value), RGB (Red, Green, Blue), etc. The electronic device may crop the image vertically to obtain a sub-image to be processed.

[0102] In one embodiment, when cropping an image, a sub-image generated within a specified time period can be cropped from the image as a sub-image to be processed. Here, the specified time period can be a time period within a specified time period before the current time point, and the specified time period can be the time period usually required for a passive target to complete a behavior in the application scenario. For example, if behaviors such as falling, running, waving, pushing, and clapping are identified, and each behavior can generally be completed within two seconds, then the specified time period is two seconds, and the specified time period is within the current two seconds.

[0103] Step 263: input the sub-image to be processed into the trained behavior recognition model to obtain the behavior recognition result output by the behavior recognition model.

[0104] After the sub-image to be processed is obtained by cutting, the sub-image to be processed can be input into a behavior recognition model for classification, and the sub-image to be processed is processed by the behavior recognition model to obtain a behavior recognition result.

[0105] Through the above measures, behavior recognition can be performed with the help of machine learning, so as to effectively process the power delay spectrum and obtain behavior recognition results.

[0106] In one embodiment, see Figure 7 , is a flow chart of a method for training a behavior recognition model provided in an embodiment of the present application, such as Figure 7 As shown, the behavior recognition model can be obtained by training through the following steps 710 to 730.

[0107] Step 710: Acquire multiple sample channel estimation information of Wi-Fi signals, and convert the multiple sample channel estimation information into sample images respectively; wherein the multiple sample channel estimation information corresponds to multiple preset behavior categories.

[0108] The sample channel estimation information is the channel estimation information collected for training the behavior recognition model. Each sample channel estimation channel corresponds to a behavior category. For example, if you want to identify the three behaviors of pushing hands, waving hands, and clapping hands, you can let people of different body shapes continue to push hands indoors (for example: do it continuously for one hour), collect the channel estimation information during this period, which is the sample channel estimation information corresponding to pushing hands; let people of different body shapes continue to wave hands indoors, collect the channel estimation information during this period, which is the sample channel estimation information corresponding to waving hands; let people of different body shapes continue to clap hands indoors, collect the channel estimation information during this period, which is the sample channel estimation information corresponding to clapping hands.

[0109] After obtaining a variety of sample channel estimation information, each sample channel estimation information is denoised and the LOS background is removed in the aforementioned manner, thereby obtaining a sample two-dimensional matrix and converting it into a sample image.

[0110] Step 720: Based on each sample image, multiple sample sub-images are obtained by cutting, and a behavior category label corresponding to the sample image is added to each sample sub-image.

[0111] After obtaining the sample images, each sample image is cut vertically to obtain multiple sample sub-images. For each sample sub-image, a behavior category label corresponding to the sample image can be added, that is, a behavior category label corresponding to the sample channel estimation information corresponding to the sample image. Based on multiple sample sub-images carrying behavior category labels, a sample data set is constructed.

[0112] Step 730: Train the machine learning model based on multiple sample sub-images carrying behavior category labels to obtain a behavior recognition model.

[0113] After obtaining the sample data set, the machine learning model is trained based on multiple sample sub-images in the sample data set. After repeated iterations, a behavior recognition model can be obtained. Here, the machine learning model can be a hybrid model of CNN (Convolutional Neural Networks), TCN (Temporal Convolutional Network), or CNN and RNN (Recurrent Neural Network). Among them, the RNN in the hybrid model can be, but is not limited to, LSTM (Long short-term memory), GRU (Gate Recurrent Unit), etc.

[0114] Figure 8is a block diagram of a behavior recognition device based on PDP according to an embodiment of the present invention. Figure 8 As shown, the device may include:

[0115] A transform module 810 is configured to transform the channel estimation information of the Wi-Fi signal into time-domain channel estimation information by inverse Fourier transform.

[0116] A decomposition module 820, configured to determine a covariance matrix of the time domain channel estimation information, and perform eigenvalue decomposition on the covariance matrix to obtain a plurality of eigenvalues;

[0117] A first determination module 830, configured to determine a target number of valid channels based on the multiple characteristic values;

[0118] A second determination module 840 is used to determine the power of multiple effective channels from the covariance matrix based on the target number;

[0119] A removal module 850, configured to remove the power of the LOS channel from the power of the multiple valid channels to obtain the power of the multiple NLOS channels;

[0120] The identification module 860 is used to perform behavior identification based on the power of the multiple NLOS channels to obtain a behavior identification result.

[0121] The implementation process of the functions and effects of each module in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned PDP-based behavior recognition method, which will not be repeated here.

[0122] In several embodiments provided in the present application, the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and a part of a module, a program segment or a code contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of boxes in the block diagram and / or the flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0123] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0124] If the function is implemented in the form of a software function module 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 application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

Claims

1. A behavior recognition method based on PDP, characterized in that: include: For the channel estimation information of the Wi-Fi signal, convert the channel estimation information into time domain channel estimation information by inverse Fourier transform; Determine a covariance matrix of the time domain channel estimation information, and perform eigenvalue decomposition on the covariance matrix to obtain a plurality of eigenvalues; Determining a target number of valid channels based on the plurality of characteristic values; Based on the target number, determining the power of a plurality of effective channels from the covariance matrix; Subtracting the power of the LOS channel from the power of the multiple effective channels to obtain the power of the multiple NLOS channels; Based on the power of the multiple NLOS channels, behavior recognition is performed to obtain a behavior recognition result; The step of determining a target number of valid channels based on the multiple characteristic values ​​comprises: Arrange the multiple eigenvalues ​​in descending order to obtain an eigenvalue sequence; wherein the eigenvalue sequence includes eigenvalues ​​numbered 0 to M; M is a positive integer; Splitting the eigenvalue sequence into a first subsequence and a second subsequence; wherein the first subsequence includes the eigenvalues ​​numbered 1 to M-1 in the eigenvalue sequence, and the second subsequence includes the eigenvalues ​​numbered 2 to M in the eigenvalue sequence; Subtracting the second subsequence from the first subsequence to obtain a difference sequence; wherein the difference sequence includes the difference values ​​from sequence number 1 to M-1; The maximum difference is found from the difference sequence, and the target number of valid channels is determined according to the sequence number of the found maximum difference.

2. The method according to claim 1, characterized in that: The determining, based on the target quantity, the powers of a plurality of effective channels from the covariance matrix comprises: Arrange multiple elements on the diagonal of the covariance matrix in order from large to small; A target number of elements at the front of the sort are selected as the power of multiple effective channels.

3. The method according to claim 1, characterized in that Before converting the channel estimation information for the Wi-Fi signal into time-domain channel estimation information by inverse Fourier transform, the method further includes: By using a boundary repetition method, the number of frequency domain channel estimation values ​​corresponding to the subcarriers in the channel estimation information is supplemented to the number of specified sampling points; A circular shift operation is performed on the complemented channel estimation information.

4. The method according to claim 1, characterized in that The performing behavior recognition based on the power of the multiple NLOS channels to obtain the behavior recognition result includes: A two-dimensional matrix is ​​constructed according to the power of multiple NLOS channels corresponding to multiple consecutive data packets; Converting the two-dimensional matrix into an image, and cutting out a sub-image to be processed from the image; The sub-image to be processed is input into a trained behavior recognition model to obtain a behavior recognition result output by the behavior recognition model.

5. The method according to claim 4, characterized in that The step of cutting out a sub-image to be processed from the image comprises: A sub-image generated within a specified time period is cut out from the image as the sub-image to be processed.

6. The method according to claim 4, characterized in that Before inputting the sub-image to be processed into a trained behavior recognition model to obtain a behavior recognition result output by the behavior recognition model, the method further includes: Acquire multiple sample channel estimation information of Wi-Fi signals, and convert the multiple sample channel estimation information into sample images respectively; wherein the multiple sample channel estimation information corresponds to multiple preset behavior categories; According to each sample image, multiple sample sub-images are obtained by cutting, and a behavior category label corresponding to the sample image is added to each sample sub-image; The machine learning model is trained based on a plurality of sample sub-images carrying behavior category labels to obtain the behavior recognition model.

7. A behavior recognition device based on PDP, characterized in that: include: A transformation module, configured to transform the channel estimation information of the Wi-Fi signal into time-domain channel estimation information by inverse Fourier transformation; A decomposition module, used to determine the covariance matrix of the time domain channel estimation information, and perform eigenvalue decomposition on the covariance matrix to obtain multiple eigenvalues; A first determination module, configured to determine a target number of valid channels based on the multiple characteristic values; A second determination module is used to determine the power of multiple effective channels from the covariance matrix based on the target number; A removal module, used for removing the power of the LOS channel from the power of the multiple effective channels to obtain the power of the multiple NLOS channels; An identification module, configured to perform behavior identification based on the power of the plurality of NLOS channels to obtain a behavior identification result; The first determination module is specifically used to: arrange the multiple eigenvalues ​​in order from large to small to obtain a eigenvalue sequence; wherein the eigenvalue sequence contains eigenvalues ​​with serial numbers 0 to M; M is a positive integer; split the eigenvalue sequence into a first subsequence and a second subsequence; wherein the first subsequence includes eigenvalues ​​with serial numbers 1 to M-1 in the eigenvalue sequence, and the second subsequence includes eigenvalues ​​with serial numbers 2 to M in the eigenvalue sequence; subtract the second subsequence from the first subsequence to obtain a differential sequence; wherein the differential sequence contains the difference values ​​from serial numbers 1 to M-1; search for the maximum difference from the differential sequence, and determine the target number of valid channels based on the serial number of the maximum difference found.

8. An electronic device, characterized in that: The electronic device comprises: processor; a memory for storing processor-executable instructions; The processor is configured to execute the PDP-based behavior recognition method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and the computer program can be executed by a processor to complete the PDP-based behavior recognition method according to any one of claims 1 to 6.

Images