In-vehicle vital sign and in-place detection system based on WiFi device
Through the in-vehicle vital signs and in-place detection system based on WiFi devices, the deep learning model is used to analyze WiFi signals, and the problems of high cost and poor environmental adaptability in the existing technology are solved, and efficient and accurate in-vehicle vital sign detection is achieved, which is suitable for cost-sensitive models.
Patent Information
- Application Number
- CN202411531376.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-30
- Publication Date
- 2025-03-11
AI Technical Summary
The existing in-vehicle wireless communication technology has problems such as high cost, poor environmental adaptability, susceptibility to interference and low detection accuracy when detecting in-vehicle vital signs and in-situation conditions.
It adopts an in-vehicle vital signs and in-place detection system based on WiFi devices, which includes a WiFi module, a DHU digital cockpit host module, an AWAS sound warning module and a TCAM vehicle communication module. The WiFi module monitors the dynamics in the vehicle in real time. The DHU module uses deep learning models to analyze signals. If a living body is detected, an alarm command will be sent to the AWAS module through the TCAM module, and the AWAS module will issue an audible alarm.
It reduces the system deployment cost, is suitable for cost-sensitive models, has strong environmental adaptability, and can be continuously monitored 24 hours a day under various weather conditions, overcomes the occlusion problem, and improves detection accuracy and application effect.
Smart Images
Figure CN119600747A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of in-vehicle wireless communication, and in particular relates to an in-vehicle vital sign and presence detection system based on WiFi equipment. Background Art
[0002] With the development of vehicle safety technology, incidents of babies being forgotten in cars are common, which may lead to serious safety consequences. Existing in-car wireless communication technologies include weight sensing, video monitoring systems, millimeter wave radar, ultra-wideband UWB technology, etc. However, these methods have many limitations. Weight sensors may not accurately detect light babies. Video monitoring systems do not work well in poor lighting conditions. Millimeter wave radars are sensitive to the environment and may be affected by various signal reflections and interference. UWB devices require the additional installation of UWB base stations and tags, have high energy consumption, high installation costs, poor convenience, and are susceptible to interference in dense electronic equipment environments.
[0003] In the field of in-vehicle vital signs and on-site detection, the OMS passenger monitoring solution based on computer vision is the closest to the existing technology. This solution uses high-definition cameras to monitor the status of people in the car in real time, and reports data in a timely manner when abnormal conditions are detected, thereby realizing the monitoring of vital signs in the car. However, the installation and maintenance costs of high-definition cameras in this computer vision-based solution are high, which is not conducive to low-cost widespread application. In addition, computer vision technology is easily affected by external factors such as lighting conditions, obstructions, and installation locations, resulting in reduced monitoring accuracy. These limitations make the applicability of existing technologies relatively limited in certain scenarios. Summary of the invention
[0004] In order to solve the above problems existing in the prior art, the present invention provides a vehicle vital sign and presence detection system based on WiFi devices. The technical problem to be solved by the present invention is achieved through the following technical solutions:
[0005] An in-vehicle vital sign and presence detection system based on WiFi devices, comprising: a WiFi module, a DHU digital cockpit host module, an AWAS sound warning module and a TCAM vehicle communication module;
[0006] The WiFi module is used to monitor the dynamics inside the vehicle in real time after the vehicle is locked to obtain an original monitoring signal; and transmit the original monitoring signal to the DHU digital cockpit host module;
[0007] The DHU digital cockpit host module is used to use the original monitoring signal and the detection algorithm to detect whether there is a living body left in the car, and if so, send an alarm instruction to the AWAS sound warning module through the TCAM vehicle communication module;
[0008] The AWAS sound warning module is used to issue a sound alarm in response to the alarm instruction;
[0009] The TCAM vehicle communication module is used to send an alarm message to the vehicle owner if no response to the alarm command is detected within a predetermined time, and to send an alarm to the police system if the life signal is still in the vehicle within the emergency set time.
[0010] Beneficial effects:
[0011] The present invention discloses a vehicle vital sign and presence detection system based on WiFi equipment. The system reuses the existing WiFi module in the vehicle to monitor the vehicle environment to obtain the original monitoring signal, and uses the deep learning model built into the DHU digital cockpit host module to analyze the original monitoring signal to confirm whether there is a life form left in the vehicle. If there is, the TCAM vehicle communication module sends an alarm instruction to the AWAS sound warning module, and the AWAS sound warning module completes the alarm. Compared with the traditional video monitoring solution, the present invention effectively reduces the system deployment cost and is particularly suitable for the promotion and application of cost-sensitive vehicle models. Wireless communication technology is adopted, which is not affected by external environmental factors such as sunlight, rain and snow, has strong environmental adaptability, can achieve 24-hour continuous monitoring under various weather conditions, and ensure the reliability of the system. It overcomes the problem of being blocked by objects such as seats and luggage in the vehicle, and can accurately detect the vital sign signals in the vehicle without being affected by obstacles, further improving the application effect and detection accuracy of the system in a complex vehicle environment.
[0012] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] Figure 1 It is a schematic diagram of an in-vehicle vital sign and presence detection system based on a WiFi device provided by the present invention;
[0014] Figure 2 It is a schematic diagram of the detection process of the in-vehicle vital sign and presence detection system based on WiFi device provided by the present invention;
[0015] Figure 3 It is a schematic diagram of the classification and recognition process of the deep learning model provided by the present invention. DETAILED DESCRIPTION
[0016] The present invention is further described in detail below with reference to specific embodiments, but the embodiments of the present invention are not limited thereto.
[0017] Combination Figure 1 and Figure 2, the present invention provides an in-vehicle vital sign and presence detection system based on WiFi equipment, including: a WiFi module, a DHU digital cockpit host module, an AWAS sound warning module and a TCAM vehicle communication module;
[0018] The WiFi module is used to monitor the dynamics inside the vehicle in real time after the vehicle is locked to obtain an original monitoring signal; and transmit the original monitoring signal to the DHU digital cockpit host module;
[0019] The WiFi module of the present invention is equipped with a transmitting antenna and a receiving antenna. The transmitting antenna uses 30 subcarriers to transmit WiFi signals multiple times per second to monitor the dynamics of the in-vehicle environment in real time to obtain original monitoring signals; wherein, when the subcarriers encounter living organisms or objects, reflection, refraction, diffraction and scattering phenomena are generated.
[0020] The WiFi module of the present invention collects carrier data more than 500 times per second and monitors the dynamics inside the vehicle in real time.
[0021] The DHU digital cockpit host module is used to use the original monitoring signal and the detection algorithm to detect whether there is a living body left in the car, and if so, send an alarm instruction to the AWAS sound warning module through the TCAM vehicle communication module;
[0022] The DHU digital cockpit host module serves as the data processing center of the system, responsible for receiving the original signal from the WiFi module and performing data preprocessing and feature extraction. This DHU digital cockpit host module is trained in conjunction with the attention feature fusion convolutional neural network model to more efficiently capture the time series features and spatial information in the signal.
[0023] The AWAS sound warning module is used to issue a sound alarm in response to the alarm instruction;
[0024] The TCAM vehicle communication module is used to send an alarm message to the vehicle owner if no response to the alarm command is detected within a predetermined time, and to send an alarm to the police system if the life signal is still in the vehicle within the emergency set time.
[0025] The TCAM module serves as the communication interface of the system. When the sound alarm is not responded to within the set time, the TCAM module will send an emergency notification to the car owner through the 4G / 5G network. It can also connect to the emergency service system for a high-level response when necessary.
[0026] After the AWAS module sounds an alarm, if the owner's response is not received within a predetermined time, the TCAM module of the present invention sends an emergency notification containing the vehicle's location, alarm information, etc. to the owner's smart device through the 4G / 5G network to ensure that the owner is aware of and takes action in a timely manner. If the owner does not respond in time, the module automatically connects to the preset emergency service system, transmits detailed emergency information, and initiates a high-level emergency response. The TCAM module uses encryption technology and security protocols to ensure data security and privacy protection during emergency notifications and communications, and prevents information from being illegally intercepted or tampered with. The TCAM module continuously monitors the status of the emergency response, uses the network to maintain communication with the owner's device and the emergency service system, provides real-time status updates and further instructions, and ensures that emergencies are handled in a timely manner.
[0027] In a specific implementation of the present invention, the WiFi module is specifically used for:
[0028] When the vehicle is locked, a WiFi signal modulated by OFDM technology is transmitted inside the vehicle;
[0029] The CSI channel state information acquisition tool is used to collect CSI data packets from the receiving antenna. The CSI data packets contain the original monitoring signals generated by the characteristic changes of the WiFi signal caused by the living body. The CSI data packets provide a rich data set for analysis.
[0030] The Wi-Fi module of the present invention can accurately collect original monitoring signals of human bodies or objects in a vehicle, and can especially monitor the left-behind conditions of infants or pets.
[0031] In a specific embodiment of the present invention, reference Figure 2 The DHU digital cockpit host module is used to receive the original monitoring signal from the WiFi module; preprocess the original monitoring signal to obtain a preprocessed signal, perform in-depth analysis on the preprocessed signal, thereby extracting detailed behavioral characteristics of the life form, and obtain an identification result of whether there is a life form in the car and the category to which the life form belongs through the trained deep learning model; if the identification result is that there is a life form in the car, an alarm instruction is sent to the AWAS sound warning module through the TCAM vehicle-mounted communication module.
[0032] The original monitoring signal X(t) is a multidimensional matrix signal including the time step, the number of subcarriers and the receiving antenna, which is expressed as:
[0033] X(t)={H t,n,m ∣t=1,2,…,T; n=1,2,…,N; m=1,2,…,M} (1)
[0034] Where: H t,n,m It represents the CSI complex value at time t, subcarrier n, and receiving antenna m; T represents the total number of time steps; N represents the number of subcarriers; and M represents the number of receiving antennas.
[0035] refer to Figure 3 The trained deep learning model of the present invention adopts a multi-channel convolutional neural network based on attention mechanism feature fusion, and the trained deep learning model includes three independent and parallel attention feature extraction convolutional neural networks, a feature loop fusion module and a Softmax classifier; the output of each attention feature extraction convolutional neural network is connected to the input of the feature loop fusion module, and the output of the feature loop fusion module is connected to the Softmax classifier; each attention feature extraction convolutional neural network includes a multi-scale convolution kernel and a fully connected layer; the output of the last convolution kernel is connected to the fully connected layer, and the output of each fully connected layer is connected to the input of the feature loop fusion module.
[0036] In a specific embodiment of the present invention, the preprocessing and in-depth analysis of the original monitoring signal to extract detailed behavioral characteristics of the living body includes:
[0037] Extract the amplitude and phase of the original monitoring signal; Amplitude A t,n,m and phase θ t,n,m Respectively expressed as:
[0038]
[0039] In the formula, Re(H t,n,m ) and Im(H t,n,m ) are H t,n,m The real and imaginary parts of
[0040] For the amplitude A t,n,m and phase θ t,n,m Normalization is performed to remove the influence of dimension and scale, and the normalized amplitude and phase are obtained, which are expressed as:
[0041]
[0042] The filter window with a window size of w is set, and the normalized amplitude and phase are median filtered to remove random noise to obtain the amplitude and phase of the median filter, which are expressed as:
[0043] A″ t,n,m =median(A t-k,n,m′ ,…,A t,n,m′ ,…,A t+k,n,m′ ) (6)
[0044] θ″t,n,m =median(θ t-k,n,m′ ,…,θ t,n,m′ ,…,θ t+k,n,m′ ) (7)
[0045] In the formula
[0046] The amplitude and phase of the median filter are filtered using an empirical mode decomposition (EMD) algorithm to obtain a preprocessed signal.
[0047] This implementation method preprocesses the collected CSI data, extracts the amplitude and phase information therein, and generates a file in CSV format for subsequent analysis.
[0048] As an optional implementation manner of the present invention, the using of the empirical mode decomposition (EMD) algorithm to filter the amplitude and phase of the median filter to obtain the preprocessed signal includes:
[0049] The median filter amplitude A″ t,n,m and phase θ″ t,n,m Decomposed into several intrinsic mode function IMF components, the amplitude and phase IMF components are expressed as:
[0050]
[0051] In the formula, IMF i (t) represents the i-th intrinsic mode function, represents the different frequency components in the signal, r(t) represents the residual signal, K represents the total number of IMF components, and L represents the total number of IMF components of the amplitude;
[0052] Select the predetermined IMF component to reconstruct the signal as needed to remove noise or interference and obtain the filtered amplitude and phase, which are expressed as:
[0053]
[0054] Where, k1, k2 represent the IMF component index range selected when reconstructing the signal, l1, l2 represent the IMF component index range selected when reconstructing the phase signal;
[0055] The filtered amplitude and phase are reorganized into CSV format data to obtain the preprocessed signal.
[0056] This implementation method extracts the amplitude and phase of the collected CSI data, normalizes it, performs median filtering and EMD algorithm filtering to generate high-quality data suitable for analysis. The deep learning model is used to deeply analyze the preprocessed signal, extract the detailed behavioral characteristics of the living organism, and enhance the model's sensitivity and recognition ability to signal changes.
[0057] In a specific embodiment of the present invention, the pre-processed signal is deeply analyzed to extract detailed behavioral characteristics of the life form, and the detailed behavioral characteristics are passed through a trained deep learning model to obtain the recognition results of whether there is a life form in the car and the category to which the life form belongs, including:
[0058] The preprocessed signal is synchronously input into three independent parallel attention feature extraction convolutional neural networks, so that the attention feature extraction convolutional neural network uses its own multi-scale convolution kernel to perform multi-level feature extraction and enhancement on the detailed behavior features, so as to obtain detailed behavior features of different scales, and output them to the feature loop fusion module through a fully connected layer;
[0059] The present invention introduces an attention mechanism into each attention feature extraction convolutional neural network, and weights the extracted features by generating query, key and value vectors to enhance the expressive power of key features. The last layer of each branch is a fully connected layer of fixed length to ensure that the output vital sign signals have consistent dimensions, which is convenient for subsequent feature fusion. The vital sign signals from the three attention feature extraction convolutional neural networks are integrated through feature fusion to form a comprehensive feature representation. The fused features are input into the Softmax classifier for classification and recognition, so as to efficiently and accurately determine whether a baby or pet is left in the car.
[0060] The feature cycle fusion module fuses the first feature input by the first fully connected layer with the second feature input by the second fully connected layer to obtain fused feature 1; fuses fused feature 1 into the first feature and the second feature to obtain the first fused feature and the second fused feature; fuses the first fused feature and the second fused feature to obtain fused feature 2; fuses fused feature 2 with the third feature input by the third fully connected layer to obtain fused feature 3; fuses fused feature 3 into fused feature 2 and the third feature to obtain the third fused feature and the fourth fused feature; fuses the third fused feature and the fourth fused feature to obtain the final fused feature, and inputs it into the Softmax classifier;
[0061] exist Figure 3 In the figure, the first to third features are represented as Feature 1 to Feature 3, and the final fused feature is represented as Fused Feature.
[0062] The Softmax classifier classifies and identifies the final fusion features to obtain an identification result of whether there is a living being in the car and the category to which the living being belongs.
[0063] This implementation uses attention features to fuse convolutional neural networks to extract detailed behavioral features, accurately extracting detailed behavioral features of living organisms, and classifying and identifying the detection results through the Softmax classification algorithm to ensure the accuracy and real-time nature of the identification.
[0064] This implementation method uses a trained deep learning model to determine whether a baby or pet is left in the car, ensuring the accuracy and real-time nature of the recognition, and providing a reliable basis for subsequent alarm triggering.
[0065] In a specific embodiment of the present invention, the AWAS sound warning module is used to respond to the alarm instruction and issue a sound alarm according to a multi-level response mechanism; wherein the multi-level response mechanism is designed according to the interval in which the variance of the detailed behavioral characteristics is located.
[0066] In this implementation, a plurality of level intervals may be designed. When the variance of the detailed behavior feature is within a corresponding level interval, a sound alarm is issued according to the corresponding level response mechanism.
[0067] For example, the module receives the alarm command from the DHU module, immediately activates the sound alarm, and emits a high-decibel alarm through the car speaker to attract the attention of the car owner or nearby people. The volume and urgency of the alarm are dynamically adjusted according to the detected life characteristics. If the baby's breathing rate is detected to be low, the system will increase the alarm intensity and extend the alarm duration, effectively attracting the attention of the car owner or surrounding people through the sound alarm, prompting them to take corresponding actions.
[0068] The present invention discloses a vehicle vital sign and presence detection system based on WiFi equipment. The system reuses the existing WiFi module in the vehicle to monitor the vehicle environment to obtain the original monitoring signal, and uses the deep learning model built into the DHU digital cockpit host module to analyze the original monitoring signal to confirm whether there is a life form left in the vehicle. If there is, the TCAM vehicle communication module sends an alarm instruction to the AWAS sound warning module, and the AWAS sound warning module completes the alarm. Compared with the traditional video monitoring solution, the present invention effectively reduces the system deployment cost and is particularly suitable for the promotion and application of cost-sensitive vehicle models. Wireless communication technology is adopted, which is not affected by external environmental factors such as sunlight, rain and snow, has strong environmental adaptability, can achieve 24-hour continuous monitoring under various weather conditions, and ensure the reliability of the system. It overcomes the problem of being blocked by objects such as seats and luggage in the vehicle, and can accurately detect the vital sign signals in the vehicle without being affected by obstacles, further improving the application effect and detection accuracy of the system in a complex vehicle environment.
[0069] It is worth noting that the terms "first" and "second" in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.
[0070] Although the present application is described herein in conjunction with various embodiments, in the process of implementing the claimed application, those skilled in the art may understand and implement other variations of the disclosed embodiments by reviewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality of components or steps.
[0071] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.
Claims
1. A vehicle vital sign and presence detection system based on WiFi equipment, characterized in that: include: WiFi module, DHU digital cockpit host module, AWAS sound warning module and TCAM vehicle communication module; The WiFi module is used to monitor the dynamics inside the vehicle in real time after the vehicle is locked to obtain the original monitoring signal; And transmit the original monitoring signal to the DHU digital cockpit host module; The DHU digital cockpit host module is used to use the original monitoring signal and the detection algorithm to detect whether there is a living body left in the car, and if so, send an alarm instruction to the AWAS sound warning module through the TCAM vehicle communication module; The AWAS sound warning module is used to issue a sound alarm in response to the alarm instruction; The TCAM vehicle communication module is used to send an alarm message to the vehicle owner if no response to the alarm command is detected within a predetermined time, and to send an alarm to the police system if the life signal is still in the vehicle within the emergency set time.
2. The in-vehicle vital signs and presence detection system based on WiFi devices according to claim 1 is characterized in that: The WiFi module is equipped with a transmitting antenna and a receiving antenna. The transmitting antenna uses 30 subcarriers to transmit WiFi signals multiple times per second to monitor the dynamics of the vehicle environment in real time to obtain the original monitoring signal; wherein, when the subcarrier encounters a living being or an object, reflection, refraction, diffraction and scattering phenomena are generated.
3. The in-vehicle vital sign and presence detection system based on WiFi device according to claim 2 is characterized in that: The WiFi module is specifically used for: When the vehicle is locked, a WiFi signal modulated by OFDM technology is transmitted inside the vehicle; The CSI channel state information acquisition tool is used to collect CSI data packets of the receiving antenna, wherein the CSI data packets contain original monitoring signals generated by characteristic changes in WiFi signals caused by living organisms.
4. The in-vehicle vital sign and presence detection system based on WiFi device according to claim 1, characterized in that: The DHU digital cockpit host module is used to receive the original monitoring signal from the WiFi module; preprocess the original monitoring signal to obtain a preprocessed signal, perform in-depth analysis on the preprocessed signal, thereby extracting detailed behavioral characteristics of the life form, and use the detailed behavioral characteristics through a trained deep learning model to obtain an identification result of whether there is a life form in the car and the category to which the life form belongs; If the recognition result is that there is a living being in the vehicle, an alarm instruction is sent to the AWAS sound warning module through the TCAM vehicle-mounted communication module.
5. The in-vehicle vital sign and presence detection system based on WiFi device according to claim 4 is characterized in that: The original monitoring signal X(t) is a multi-dimensional matrix signal including the time step, the number of subcarriers and the receiving antenna, which is expressed as: X(t)={H t,n,m ∣t=1,2,…,T;n=1,2,…,N;m=1,2,…,M} (1) Where: H t,n,m It represents the CSI complex value at time t, subcarrier n, and receiving antenna m; T represents the total number of time steps; N represents the number of subcarriers; and M represents the number of receiving antennas.
6. The in-vehicle vital sign and presence detection system based on WiFi device according to claim 5, characterized in that: The preprocessing of the original monitoring signal comprises: Extract the amplitude and phase of the original monitoring signal; Amplitude A t,n,m and phase θ t,n,m Respectively expressed as: In the formula, Re(H t,n,m ) and Im(H t,n,m ) are H t,n,m The real and imaginary parts of For the amplitude A t,n,m and phase θ t,n,m Normalization is performed to remove the influence of dimension and scale, and the normalized amplitude and phase are obtained, which are expressed as: The filter window with a window size of w is set, and the normalized amplitude and phase are median filtered to remove random noise to obtain the amplitude and phase of the median filter, which are expressed as: A″ t,n,m =median(A t-k,n,m′ ,…,A t,n,m′ ,…,A t+k,n,m′ ) (6) θ″ t,n,m =median(θ t-k,n,m′ ,…,θ t,n,m′ ,…,θ t+k,n,m′ ) (7) In the formula The amplitude and phase of the median filter are filtered using an empirical mode decomposition (EMD) algorithm to obtain a preprocessed signal.
7. The in-vehicle vital sign and presence detection system based on WiFi device according to claim 6, characterized in that: The method of using the empirical mode decomposition (EMD) algorithm to filter the amplitude and phase of the median filter to obtain the preprocessed signal includes: The median filter amplitude A″ t,n,m and phase θ″ t,n,m Decomposed into several intrinsic mode function IMF components, the amplitude and phase IMF components are expressed as: In the formula, IMF i (t) represents the i-th intrinsic mode function, represents the different frequency components in the signal, r(t) represents the residual signal, K represents the total number of IMF components, and L represents the total number of IMF components of the amplitude; Select the predetermined IMF component to reconstruct the signal as needed to remove noise or interference and obtain the filtered amplitude and phase, which are expressed as: Where, k1, k2 represent the IMF component index range selected when reconstructing the signal, l1, l2 represent the IMF component index range selected when reconstructing the phase signal; The filtered amplitude and phase are reorganized into CSV format data to obtain the preprocessed signal.
8. The in-vehicle vital sign and presence detection system based on WiFi device according to claim 5, characterized in that: The trained deep learning model is a multi-channel convolutional neural network based on attention mechanism feature fusion, and the trained deep learning model includes three independent and parallel attention feature extraction convolutional neural networks, a feature loop fusion module and a Softmax classifier; the output of each attention feature extraction convolutional neural network is connected to the input of the feature loop fusion module, and the output of the feature loop fusion module is connected to the Softmax classifier; each attention feature extraction convolutional neural network includes a multi-scale convolution kernel and a fully connected layer; the output of the last convolution kernel is connected to the fully connected layer, and the output of each fully connected layer is connected to the input of the feature loop fusion module.
9. The in-vehicle vital sign and presence detection system based on WiFi device according to claim 8, characterized in that: The pre-processed signal is subjected to a deep analysis to extract detailed behavioral features of the life form, and the detailed behavioral features are subjected to a trained deep learning model to obtain an identification result of whether there is a life form in the vehicle and the category to which the life form belongs, including: The preprocessed signal is synchronously input into three independent parallel attention feature extraction convolutional neural networks, so that the attention feature extraction convolutional neural network uses its own multi-scale convolution kernel to perform multi-level feature extraction and enhancement on the detailed behavior features, so as to obtain detailed behavior features of different scales, and output them to the feature loop fusion module through a fully connected layer; The feature cycle fusion module fuses the first feature input by the first fully connected layer with the second feature input by the second fully connected layer to obtain fused feature 1; fuses fused feature 1 into the first feature and the second feature to obtain the first fused feature and the second fused feature; fuses the first fused feature and the second fused feature to obtain fused feature 2; fuses fused feature 2 with the third feature input by the third fully connected layer to obtain fused feature 3; fuses fused feature 3 into fused feature 2 and the third feature to obtain the third fused feature and the fourth fused feature; fuses the third fused feature and the fourth fused feature to obtain the final fused feature, and inputs it into the Softmax classifier; The Softmax classifier classifies and identifies the final fusion features to obtain an identification result of whether there is a living being in the car and the category to which the living being belongs.
10. The in-vehicle vital sign and presence detection system based on WiFi device according to claim 4, characterized in that: The AWAS sound warning module is used to respond to the warning instruction and issue a sound alarm according to a multi-level response mechanism; wherein the multi-level response mechanism is designed according to the interval in which the variance of the detailed behavior characteristics is located.