Human body behavior recognition method and device, electronic equipment and storage medium
By acquiring radar reflection signals and using a trained three-channel convolutional neural network for human behavior recognition, combined with behavior change verification, the final classification result solves the problem of high sample dependence in existing technologies and improves recognition accuracy and robustness.
Patent Information
- Application Number
- CN202511425125.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-30
- Publication Date
- 2026-02-10
AI Technical Summary
Existing radar- and deep learning-based methods for human behavior recognition are highly dependent on samples, and are prone to recognition bias, especially when there are insufficient training samples, particularly in state transition behaviors such as falls.
By acquiring human body reflection signals collected by radar, the changes in behavior are determined. A pre-trained three-channel convolutional neural network based on an attention mechanism is used for preliminary classification, and the final classification results are verified based on the changes, reducing the reliance on large-scale fall samples.
It improves the accuracy and robustness of human behavior recognition, especially in recognizing state transition behaviors such as falling and getting up, and overcomes the problem of over-reliance on a single feature or a large number of fall samples.
Smart Images
Figure CN121502441A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of radar technology, and more particularly to a human behavior recognition method, device, electronic device, and storage medium. Background Technology
[0002] It should be noted that the above description of the technical background is only for the purpose of providing a clear and complete explanation of the technical solutions of the present invention and facilitating understanding by those skilled in the art. It should not be assumed that the above technical solutions are known to those skilled in the art simply because they have been described in the background section of this invention.
[0003] Accurate identification and classification of human behavior are of great significance in scenarios such as smart healthcare, intelligent security, elderly care monitoring, and human-computer interaction. Radar, with its advantages of being non-contact, all-weather, and able to penetrate obstructions, has attracted much attention.
[0004] Related technologies often combine radar technology with depth technology to achieve human behavior monitoring through feature extraction and classification. However, this method is highly dependent on samples, and recognition bias can easily occur when there are insufficient training samples. Summary of the Invention
[0005] In view of the above, the purpose of one or more embodiments of this disclosure is to provide a human behavior recognition method, device, electronic device and storage medium to solve the problems of the prior art.
[0006] To achieve the above objectives, one or more embodiments of this disclosure provide a method for human behavior recognition, including:
[0007] Acquire human body reflection signals collected by radar;
[0008] Based on the human body reflection signals, the changes in human behavior are determined;
[0009] Based on the human body reflection signals, a preliminary classification result is obtained through a trained human behavior classification model;
[0010] In response to the constraint that the preliminary classification result satisfies the change scenario, the final classification result of human behavior is determined.
[0011] Optionally, after acquiring the human body reflection signal, the method further includes:
[0012] The human body reflection signal is subjected to mean cancellation processing to obtain the filtered human body reflection signal;
[0013] For each frame of the human body reflection signal, a two-dimensional fast Fourier transform is performed to obtain the distance and velocity information of the human body reflection signal;
[0014] Based on the distance information and the velocity information, construct a distance-time matrix, a velocity-time matrix, and a distance-velocity matrix;
[0015] The distance-time matrix, the velocity-time matrix, and the distance-velocity matrix are converted into a thermodynamic spectrum.
[0016] Optionally, determining the changes in human behavior based on the human body reflection signals includes:
[0017] Statistically analyze the power burst curves of the aforementioned velocity-time matrix;
[0018] Based on the power burst curve and the preset threshold, determine whether the human behavior includes a static state;
[0019] The state of the human body behavior includes a stationary state, and the stationary posture of the human body behavior is determined according to the distance-velocity matrix.
[0020] Optionally, based on the power burst curve and a preset threshold, determining whether the human behavior includes a static state includes:
[0021] Based on the velocity-time matrix, the positive and negative motion components of the Doppler spectrum intensity generated by human motion at different times are obtained.
[0022] Based on the positive motion vector and the negative motion component, the total intensity of the Doppler spectrum generated by human motion at different times is obtained;
[0023] Based on the total intensity of the Doppler spectrum and a preset threshold, the state of human behavior at different times is determined, including a static state and a dynamic state.
[0024] Optionally, determining the stationary posture of the human body behavior based on the distance-velocity matrix includes:
[0025] Based on the distance-velocity matrix, the energy distribution curve is obtained;
[0026] Determine the average fluctuation length of the energy distribution curve under static conditions;
[0027] The static posture of the human body is determined based on the fact that the average fluctuation length is less than a preset threshold. The static posture includes standing posture and lying posture.
[0028] Optionally, based on the human body reflection signal, a preliminary classification result is obtained using a trained human behavior classification model, including:
[0029] The heat map is input into the human behavior classification model to obtain the preliminary classification result. The human classification model is trained on a three-channel convolutional neural network based on an attention mechanism.
[0030] Optionally, the final classification result of human behavior is determined, including:
[0031] The behavioral logic of the human body behavior is determined based on the movement state and / or static posture of the human body behavior.
[0032] In response to determining that the preliminary classification result satisfies the behavioral logic, the final classification result of the human behavior is determined.
[0033] Based on the same inventive concept, one or more embodiments of this disclosure also provide a human behavior recognition device, including:
[0034] The acquisition module is configured to acquire human body reflection signals collected by radar;
[0035] The first determining module is configured to determine the changes in human behavior based on the human body reflection signals.
[0036] The second determining module is configured to obtain a preliminary classification result based on the human body reflection signal and a trained human behavior classification model.
[0037] The classification module is configured to determine the final classification result of human behavior in response to the constraint that the preliminary classification result meets the changing conditions.
[0038] Based on the same inventive concept, one or more embodiments of this disclosure also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the human behavior recognition method as described in any of the above.
[0039] Based on the same inventive concept, one or more embodiments of this disclosure also provide a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute any of the human behavior recognition methods described above.
[0040] As can be seen from the above, the human behavior recognition method provided by one or more embodiments of this disclosure first determines the changes in human behavior based on human reflection signals, uses the changes as constraints on the output results of the human behavior classification model, and further judges whether the output results conform to human behavior logic, thereby improving the accuracy and stability of human behavior recognition.
[0041] The human behavior recognition device, electronic device, and computer-readable storage medium disclosed herein can all implement the steps of the above-described human behavior recognition method, and therefore also possess the beneficial effects of the above-described human behavior recognition method. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in one or more embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only one or more embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0043] Figure 1 This is a flowchart illustrating one or more embodiments of the human behavior recognition method disclosed herein;
[0044] Figure 2 This is a thermogram of one embodiment of the present disclosure;
[0045] Figure 3 This is a power burst curve of one embodiment of the present disclosure;
[0046] Figure 4 An energy curve of one embodiment of this disclosure;
[0047] Figure 5 This is a schematic diagram of the structure of a human behavior recognition device according to one or more embodiments of the present disclosure;
[0048] Figure 6 This is a schematic diagram of the hardware structure of an electronic device according to one or more embodiments of this disclosure. Detailed Implementation
[0049] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.
[0050] It should be noted that, unless otherwise defined, the technical or scientific terms used in one or more embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar words used in one or more embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0051] As mentioned in the background section, how to improve the accuracy and robustness of human behavior classification has always been a hot research topic.
[0052] Traditional methods for human behavior recognition primarily rely on visual sensors, wearable devices, and wireless signal sensing. However, while visual sensors can provide rich image information, they are limited by lighting conditions and privacy concerns; wearable devices can directly collect motion information, but user compliance is low, and they are not suitable for long-term continuous monitoring.
[0053] Therefore, related technologies have proposed radar-based behavior recognition. However, current behavior recognition methods based on frequency-modulated continuous wave radar mainly combine radar with deep learning technology to achieve human behavior recognition through feature extraction and classification. However, this method is highly dependent on samples, especially in state transition behaviors such as falls, where insufficient training samples can easily lead to recognition bias.
[0054] Based on some implementations of this disclosure, a scheme for human behavior recognition is provided. In this scheme, firstly, human body reflection signals collected by radar are acquired; based on the human body reflection signals, the changing patterns of human behavior are determined; then, based on the human body reflection signals, a trained human behavior classification model is used to obtain a preliminary classification result; finally, in response to the constraint that the preliminary classification result satisfies the changing patterns, the final classification result of the human behavior is determined. Through the scheme of this disclosure, the preliminary classification result can be verified by combining the actual changes in human behavior, effectively improving the accuracy and reliability of human behavior recognition.
[0055] refer to Figure 1 The present disclosure discloses a human behavior recognition method according to one or more embodiments, including the following steps:
[0056] Step S101: Acquire the human body reflection signal collected by the radar;
[0057] Step S102: Determine the changes in human behavior based on the aforementioned human reflex signals;
[0058] Step S103: Based on the above human body reflection signals, a preliminary classification result is obtained through the trained human behavior classification model;
[0059] Step S104: In response to the constraints of the above preliminary classification results satisfying the above-mentioned change conditions, determine the final classification result of human behavior.
[0060] In the implementation of this disclosure, the raw echo signal collected by radar is used as the data basis for human behavior recognition and analysis.
[0061] In the embodiments of this disclosure, a radar simulation system based on Frequency Modulated Continuous Wave Radar (FMCW) technology can be used to simulate and verify the impact of low pulse repetition frequency on respiratory and heartbeat extraction. For example, a Ti6843 radar module can be used to collect human behavior data.
[0062] In the embodiments disclosed herein, the radar sweep frequency can be designed to be 60 GHz, the sweep slope to be 57 GHz / ms, the number of sampling points of the analog-to-digital converter (ADC) to be 128, the ADC data acquisition rate to be 4 MHz, the single frame duration to be 0.039 s, the system bandwidth to be 3.657 GHz, and the range resolution to be 0.04 m. The radar can be mounted at a distance of 1-5 m from the human body, at a height of 0.8 m, and placed perpendicular to the ground.
[0063] In the implementation of this disclosure, after acquiring the human body reflection signal, stationary clutter can be filtered out through mean cancellation processing. For example, data can be acquired from various human behaviors, with one hundred frames of data acquired for each behavior. Stationary clutter can be filtered out by performing mean cancellation on each frame of data.
[0064] In the implementation of this disclosure, determining the changes in human behavior and preliminary classification results based on human body reflection signals may include: obtaining distance and speed information of the human body; obtaining the changes in human behavior based on the distance and speed information; and obtaining preliminary classification results through a human behavior classification model.
[0065] In the implementation of this disclosure, a Fast Fourier Transform (FFT) is performed on each frame of data to obtain distance and velocity information.
[0066] In the implementation of this disclosure, the calculation formula can be:
[0067]
[0068] Where r represents distance, v represents speed, N represents the number of fast-time sampling points, M represents the number of slow-time sampling points, s(n,m) represents the echo signal after mean elimination, and j represents the imaginary unit.
[0069] The distance and velocity information is stitched together frame by frame to obtain the distance-time matrix, the velocity-time matrix, and the distance-velocity matrix obtained by superimposing the distance-time matrix and the velocity-time matrix frame by frame. These three types of matrices are then converted into a heat map. It can be understood that the distance-time matrix, the velocity-time matrix, and the distance-velocity matrix reflect the time-frequency characteristics of the human body's reflected signals.
[0070] Figure 2 The thermal spectrum obtained in one embodiment of this disclosure shows the relationship between time and distance, time and velocity, and distance and velocity from different dimensions. This thermal spectrum can be used as input data for a subsequent three-channel human body classification model.
[0071] In the implementation of this disclosure, determining the change in human behavior based on the human body reflection signal may include: dividing human behavior into a static state and an action state based on a velocity-time matrix; and further calculating the static state to determine the static posture.
[0072] In the implementation of this disclosure, the process of dividing human behavior into a static state and an active state may include: the power-burst curve (PBC) of the statistical velocity-time matrix, wherein the formula for calculating PBC is: PBC(t) = PBC + (t)+PBC - (t), S(t,v) represents the velocity at time t, PBC + (t) represents positive motion (motion away from the radar), PBC - (t) represents negative motion (motion closer to the radar); it is determined whether the PBC value corresponding to each time t exceeds the preset threshold. If it is less than the threshold, it is determined that the time is in motion state. If it is greater than or equal to the threshold, it is determined that the time is in stationary state.
[0073] Figure 3 The diagram shows a power burst curve according to an embodiment of this disclosure, including a positive motion curve and a negative motion curve. By studying the positive and negative motion curves, the entire human body behavior can be divided into multiple segments, each representing a motion state or a static state.
[0074] In the implementation of this disclosure, the process of further calculating the stationary state to determine the stationary attitude may include: obtaining the range-dimensional energy distribution curve within each frame based on the range-velocity matrix, wherein the energy distribution calculation formula is: The average fluctuation length of the energy curve within the static interval is calculated using the following formula: When L avg When the value exceeds a preset threshold, the static posture is determined to be lying down. avg When the value is less than a preset threshold, the static posture is judged as standing.
[0075] Figure 4 The energy distribution curve of one embodiment of this disclosure shows the relationship between energy and distance.
[0076] In the implementation of this disclosure, behavioral logic is determined based on the motion state and the stationary state, as well as the posture in the stationary state.
[0077] In the embodiments of this disclosure, when the behavior state is "pre-stationary state-movement state-post-stationary state" or "pre-stationary state-post-stationary state", and the pre-stationary state is standing and the post-stationary state is lying down, the behavior logic is falling down.
[0078] In the implementation of this disclosure, a preliminary classification result is obtained based on the human body reflection signal and a trained human behavior classification model. This may include: inputting a thermal spectrum into the human behavior classification model and obtaining the output preliminary classification result.
[0079] In the implementation of this disclosure, the above-mentioned human behavior classification model can be trained using a three-channel convolutional neural network based on an attention mechanism.
[0080] In the implementation disclosed herein, the cross-entropy loss function of the human behavior classification model can be:
[0081]
[0082] Where K represents the number of human behavior categories, y k This represents the output of the human behavior classification model.
[0083] In the implementation of this disclosure, it is necessary to determine whether the category output by the human body classification model satisfies the change scenario of human behavior.
[0084] In the embodiments of this disclosure, when the category output by the human body classification model is "falling" and the logic of the change in human behavior is also "falling", the final category is determined to be "falling".
[0085] In the embodiments of this disclosure, when the category output by the human body classification model is "standing" but the logic of the change in human behavior is "falling", the classification is determined to be a failure and the classification is re-performed.
[0086] This disclosure reduces the reliance of deep learning models on large-scale fall samples, improving the accuracy and robustness of recognizing state transition behaviors such as falling and getting up. In other words, this disclosure effectively overcomes the problem of over-reliance on single features or a large number of fall samples in existing technologies, achieving high-precision recognition and improved robustness of human behavior in complex scenarios, especially showing significant advantages in monitoring state transition behaviors such as falling and getting up.
[0087] It is understandable that this method can be executed by any device, equipment, platform, or cluster of devices with computing and processing capabilities.
[0088] It should be noted that the methods of one or more embodiments of this disclosure can be executed by a single device, such as a computer or server. The methods of this embodiment can also be applied in a distributed scenario, where multiple devices cooperate to complete the process. In such a distributed scenario, one of these devices may execute only one or more steps of the methods of one or more embodiments of this disclosure, and the multiple devices will interact with each other to complete the method described.
[0089] It should be noted that the above description pertains to specific embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than those shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
[0090] Based on the same inventive concept, corresponding to any of the methods in the above embodiments, this disclosure also provides a human behavior recognition device. For example... Figure 5 As shown, the device includes:
[0091] The acquisition module 11 is configured to acquire human body reflection signals collected by the radar;
[0092] The first determining module 12 is configured to determine the changes in human behavior based on the human body reflection signal;
[0093] The second determining module 13 is configured to obtain a preliminary classification result based on the human body reflection signal and a trained human behavior classification model.
[0094] The classification module 14 is configured to determine the final classification result of human behavior in response to the constraint that the preliminary classification result meets the changing conditions.
[0095] Optionally, it is also configured as follows:
[0096] After acquiring the human body reflection signal, the method further includes:
[0097] The human body reflection signal is subjected to mean cancellation processing to obtain the filtered human body reflection signal;
[0098] For each frame of the human body reflection signal, a two-dimensional fast Fourier transform is performed to obtain the distance and velocity information of the human body reflection signal;
[0099] Based on the distance information and the velocity information, construct a distance-time matrix, a velocity-time matrix, and a distance-velocity matrix;
[0100] The distance-time matrix, the velocity-time matrix, and the distance-velocity matrix are converted into a thermodynamic spectrum.
[0101] Optionally, the first determining module 12 is specifically configured as follows:
[0102] Statistically analyze the PBC curve of the velocity-time matrix;
[0103] Based on the PBC curve and the preset threshold, determine whether the human behavior includes a static state;
[0104] The state of the human body behavior includes a stationary state, and the stationary posture of the human body behavior is determined according to the distance-velocity matrix.
[0105] Optionally, the first determining module 12 is specifically configured as follows:
[0106] Based on the velocity-time matrix, the positive and negative motion components of the Doppler spectrum intensity generated by human motion at different times are obtained.
[0107] Based on the positive motion vector and the negative motion component, the total intensity of the Doppler spectrum generated by human motion at different times is obtained;
[0108] Based on the total intensity of the Doppler spectrum and a preset threshold, the state of human behavior at different times is determined, including a static state and a dynamic state.
[0109] Optionally, the first determining module 12 is specifically configured as follows:
[0110] Based on the distance-velocity matrix, the energy distribution curve is obtained;
[0111] Determine the average fluctuation length of the energy distribution curve under static conditions;
[0112] The static posture of the human body is determined based on the fact that the average fluctuation length is less than a preset threshold. The static posture includes standing posture and lying posture.
[0113] Optionally, the second determining module 13 is specifically configured as follows:
[0114] The heat map is input into the human behavior classification model to obtain the preliminary classification result. The human classification model is trained on a three-channel convolutional neural network based on an attention mechanism.
[0115] Optionally, the classification module 14 is specifically configured as follows:
[0116] The behavioral logic of the human body behavior is determined based on the movement state and / or static posture of the human body behavior.
[0117] In response to determining that the preliminary classification result satisfies the behavioral logic, the final classification result of the human behavior is determined.
[0118] For ease of description, the above apparatus is described in terms of its functions, divided into various modules. Of course, when implementing one or more embodiments of this disclosure, the functions of each module can be implemented in one or more software and / or hardware.
[0119] The apparatus described above is used to implement the corresponding methods in the foregoing embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0120] Figure 6 This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.
[0121] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this disclosure.
[0122] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this disclosure are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.
[0123] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.
[0124] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).
[0125] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.
[0126] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this disclosure, and not necessarily all the components shown in the figures.
[0127] The electronic devices described above are used to implement the corresponding methods in the foregoing embodiments and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0128] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.
[0129] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of this disclosure as described above, which are not provided in detail for the sake of brevity.
[0130] Additionally, to simplify the description and discussion, and to avoid obscuring one or more embodiments of this disclosure, the provided drawings may or may not show well-known power / ground connections to integrated circuit (IC) chips and other components. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring one or more embodiments of this disclosure, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which one or more embodiments of this disclosure will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuitry) are set forth to describe exemplary embodiments of this disclosure, it will be apparent to those skilled in the art that one or more embodiments of this disclosure may be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.
[0131] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.
[0132] This disclosure includes one or more embodiments intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments of this disclosure should be included within the scope of protection of this disclosure.
Claims
1. A method for recognizing human behavior, characterized in that, include: Acquire human body reflection signals collected by radar; Based on the human body reflection signals, the changes in human behavior are determined; Based on the human body reflection signals, a preliminary classification result is obtained through a trained human behavior classification model; In response to the constraint that the preliminary classification result satisfies the change scenario, the final classification result of human behavior is determined.
2. The method according to claim 1, characterized in that, After acquiring the human body reflection signal, the method further includes: The human body reflection signal is subjected to mean cancellation processing to obtain the filtered human body reflection signal; For each frame of the human body reflection signal, a two-dimensional fast Fourier transform is performed to obtain the distance and velocity information of the human body reflection signal; Based on the distance information and the velocity information, construct a distance-time matrix, a velocity-time matrix, and a distance-velocity matrix; The distance-time matrix, the velocity-time matrix, and the distance-velocity matrix are converted into a thermodynamic spectrum.
3. The method according to claim 2, characterized in that, Based on the human body reflection signals, determine the changes in human behavior, including: Statistically analyze the power burst curves of the aforementioned velocity-time matrix; Based on the power burst curve and the preset threshold, determine whether the human behavior includes a static state; The state of the human body behavior includes a stationary state, and the stationary posture of the human body behavior is determined according to the distance-velocity matrix.
4. The method according to claim 3, characterized in that, Based on the power burst curve and a preset threshold, it is determined whether the human behavior includes a static state, including: Based on the velocity-time matrix, the positive and negative motion components of the Doppler spectrum intensity generated by human motion at different times are obtained. Based on the positive motion vector and the negative motion component, the total intensity of the Doppler spectrum generated by human motion at different times is obtained; Based on the total intensity of the Doppler spectrum and a preset threshold, the state of human behavior at different times is determined, including a static state and a dynamic state.
5. The method according to claim 3, characterized in that, Determining the static posture of the human body based on the distance-velocity matrix includes: Based on the distance-velocity matrix, the energy distribution curve is obtained; Determine the average fluctuation length of the energy distribution curve under static conditions; The static posture of the human body is determined based on the fact that the average fluctuation length is less than a preset threshold. The static posture includes standing posture and lying posture.
6. The method according to claim 2, characterized in that, Based on the human body reflection signals, a preliminary classification result is obtained using a trained human behavior classification model, including: The heat map is input into the human behavior classification model to obtain the preliminary classification result. The human classification model is trained on a three-channel convolutional neural network based on an attention mechanism.
7. The method according to claim 3, characterized in that, The final classification of human behavior is determined by: The behavioral logic of the human body behavior is determined based on the movement state and / or static posture of the human body behavior. In response to determining that the preliminary classification result satisfies the behavioral logic, the final classification result of the human behavior is determined.
8. A human behavior recognition device, characterized in that, include: The acquisition module is configured to acquire human body reflection signals collected by radar; The first determining module is configured to determine the changes in human behavior based on the human body reflection signals. The second determining module is configured to obtain a preliminary classification result based on the human body reflection signal and a trained human behavior classification model. The classification module is configured to determine the final classification result of human behavior in response to the constraint that the preliminary classification result meets the changing conditions.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executed by the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 7.
10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1 to 7.