Multi-unit radar coordinated motion state and vital sign monitoring system and method

Abstract

The application discloses a kind of multi-unit radar coordination's motion state and vital sign monitoring system and method.The system includes the edge sensing node and host computer of communication connection;Edge sensing node has complete vital sign wireless sensing, signal solution and communication ability;Host computer is the target ID, target position, target vital sign and target abnormal state etc. information sent by receiving edge sensing node;The edge sensing node is composed of microcontroller module, radar sensor module, DSP module, FPGA neural network inference module, communication module and power module;Microcontroller module is connected with radar sensor module, DSP module, FPGA neural network inference module and communication module communication respectively;Radar sensor module is connected with DSP module communication;DSP module and FPGA neural network inference module are connected with communication;Power module is the power supply of edge sensing node.The method carries out the accurate tracking of multi-target vital sign and position, realizes the accurate tracking of multi-target health state.

Description

Technical Field

[0001] This invention relates to the field of wireless sensing technology, specifically to a multi-unit radar collaborative motion state and vital sign monitoring system and method. Background Technology

[0002] With the accelerating aging of society, the demand for real-time health monitoring and assessment technologies in nursing homes and other centralized elderly care facilities is becoming increasingly prominent. Vital signs (such as heartbeat and respiration) and behavioral status (such as falls) of the elderly are important indicators for assessing health status and preventing accidents. Long-term, accurate monitoring of these indicators is crucial for timely detection of potential health risks and improving care efficiency. Traditional contact-based monitoring technologies (such as electrocardiograms and photoplethysmography) rely on patches, sensors, or cable connections, which can easily restrict daily activities, resulting in poor wearing comfort and failing to meet the needs of all-weather, multi-target health monitoring. Regarding non-contact technologies, vision-based measurement methods, while avoiding wearing restrictions, require a light source and are easily affected by the field of view and ambient lighting, while also posing a certain risk of privacy breaches, making widespread adoption in nursing homes, homes, and other similar settings difficult.

[0003] Radar-based vital sign monitoring technology has become an important research direction in the field of health monitoring due to its non-invasive, non-contact, non-line-of-sight, multi-target simultaneous measurement, and privacy protection characteristics. However, existing single-node radar monitoring systems are susceptible to multipath effects and obstruction interference in multi-target scenarios, making it difficult to achieve accurate coverage of multiple areas and simultaneous monitoring of multiple objects. In summary, due to the rich information and large data volume generated by radar high-frequency signals, current systems still face challenges in data processing efficiency when dealing with large-area, multi-target human health status monitoring; electromagnetic wave signals are easily obstructed and subject to multipath interference, resulting in insufficient stability of continuous measurement of human health parameters in complex environments. In addition, the visualization and analysis of human health status parameters, as well as anomaly identification and alarm functions, still need improvement.

[0004] The paper "Ding Yi. Development and Implementation of Indoor and Outdoor Multi-Person Vital Sign Monitoring System Based on Multi-Radar Fusion IoT [D]. Beijing University of Posts and Telecommunications, 2022" describes the development of a multi-radar fusion IoT system with three ultra-wideband radars and one continuous wave radar, achieving motion attitude monitoring and vital sign monitoring of multiple moving targets. However, this system uses multiple nodes within a single room for measurement and does not extend to multiple rooms with interconnected nodes to achieve wider-area monitoring. Therefore, there is an urgent need to design a fully functional monitoring system and corresponding monitoring methods. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the technical problem this invention aims to solve is to provide a multi-unit radar collaborative motion status and vital signs monitoring system and method.

[0006] The technical solution of the present invention to solve the aforementioned system technical problem is to provide a multi-unit radar collaborative motion state and vital signs monitoring system, characterized in that the system includes an edge sensing node and a host computer with communication connection;

[0007] The edge sensing node has complete capabilities for wireless sensing of vital signs, signal processing, and communication; the host computer receives information such as target ID, target location, target vital signs, and target abnormal status sent by the edge sensing node;

[0008] The edge sensing node consists of a microcontroller module, a radar sensor module, a DSP module, an FPGA neural network inference module, a communication module, and a power supply module. The microcontroller module is communicatively connected to the radar sensor module, the DSP module, the FPGA neural network inference module, and the communication module. The radar sensor module is communicatively connected to the DSP module. The DSP module and the FPGA neural network inference module are communicatively connected. The power supply module provides power to the edge sensing node.

[0009] The technical solution of this invention to solve the aforementioned method problem is to provide a method for monitoring motion status and vital signs using multi-unit radar coordination, characterized in that the method includes the following steps:

[0010] Step 1: Set up a monitoring system;

[0011] Step 2: First, the radar sensor module dynamically adjusts the measurement parameters according to the instructions issued by the microcontroller module, transmits signals and receives echo signals; then, the radar sensor module completes target information acquisition and obtains a three-dimensional data matrix composed of raw ADC data; subsequently, the DSP module outputs a three-dimensional feature dataset containing target distance, angle and velocity.

[0012] Step 3: Input the three-dimensional feature dataset generated in Step 2 into the DSP module to perform multi-level preprocessing and output the primary target motion trajectory; then, based on the angle information in the primary target motion trajectory, perform beamforming based on the minimum variance distortionless response algorithm to enhance the target direction signal and form a primary four-dimensional feature dataset containing the displacement information of the target thoracic cavity surface.

[0013] Step 4: The microcontroller module performs multi-dimensional analysis on the primary four-dimensional feature dataset obtained in Step 3 to establish a target occlusion processing and monitoring relay mechanism; then, based on the target occlusion processing and monitoring relay mechanism, it processes the primary target motion trajectory and outputs a stable and continuous four-dimensional feature dataset.

[0014] Step 5: Extraction of heartbeat and breathing features and fall detection features;

[0015] In the heartbeat and respiration feature extraction, firstly, based on the target chest cavity micro-motion features in the four-dimensional feature dataset output in step 4, variational mode decomposition is performed in the DSP module to obtain several intrinsic mode functions. In the decomposition results, if there are mode functions with energy concentrated in the 0.1-0.8Hz respiratory frequency band or the 1-3Hz heartbeat frequency band, and the frequency domain sidelobe suppression ratio is ≥20dB, they are determined to be valid vital sign signals; otherwise, they are determined to be low signal-to-noise ratio signals that do not meet the conditions, and are input into the FPGA neural network inference module for deep feature mining to obtain vital sign parameters. Then, target action influence correction is performed to obtain stable vital sign signals.

[0016] In the fall detection feature extraction, the four-dimensional feature dataset output in step 4 is used by the DSP module to perform differential operations to extract the vertical velocity component of the human body's center of mass, and a two-layer detection mechanism based on kinematic criteria is constructed to determine whether a fall event has occurred.

[0017] Step 6: Transmit the vital signs signals and whether a fall event has occurred obtained in Step 5 to the host computer for display and statistics.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0019] 1. Construct a multi-node edge sensing system to achieve wide-area vital sign monitoring;

[0020] This invention constructs a multi-node edge sensing system based on millimeter-wave radar, enabling 24 / 7 wide-area monitoring and analysis of human health and abnormal conditions in locations such as nursing homes. Its technical solution efficiently addresses the problems of signal loss and false detection in multi-target monitoring caused by factors such as obstruction and human activity in complex environments. Furthermore, by optimizing resource deployment through node-side edge computing, it solves the problem of insufficient computing resources, reduces the pressure of remote data transmission, and significantly improves the real-time performance and reliability of monitoring.

[0021] 2. A motion interference suppression and occlusion recovery recognition compensation algorithm is proposed for accurate tracking of multiple target vital signs and locations. The proposed interference suppression and compensation algorithm includes:

[0022] (1) Based on the short-term invariance of vital signs, target ID, target location, and vital sign information are bound together. Specifically, the target number, historical location, predicted location based on extended Kalman filter, and vital sign similarity are bound together. In the case of target recovery after occlusion, the similarity between the new target location and the predicted location group based on Kalman filter, and between the new target vital signs and the target vital signs group before occlusion, are evaluated to achieve target re-association, which significantly improves the target recognition accuracy in the case of occlusion recovery and avoids target loss caused by interference;

[0023] (2) Construct a vital sign regression algorithm based on historical vital sign information. If the current detection value is an outlier of historical vital signs, it is determined to be a false value caused by interference and is regressed to the baseline value.

[0024] (3) When vital signs cannot be detected at all, Kalman filtering is used for short-term prediction, and the prediction method is iteratively optimized by combining the actual detection values ​​after the subsequent interference disappears, so as to ensure data continuity and accuracy.

[0025] 3. An adaptive network communication structure and node linkage function based on WiFi Mesh are proposed;

[0026] This invention constructs an edge sensing node network based on WiFi Mesh, enabling efficient communication between nodes. The network is adaptive, allowing the system to continue operating normally even when some nodes are unavailable, giving the system advantages such as flexibility, scalability, and node linkage.

[0027] 4. A task switching algorithm based on multi-nodes and map models is proposed to achieve wide-area uninterrupted monitoring;

[0028] By deploying multiple radar nodes in multiple rooms, when a target leaves the field of view of the current node, the current node can automatically transfer the monitoring task and target information to a new node in the room where the target enters, based on the target's movement trend and the pre-input map model. This enables uninterrupted and wide-area monitoring in scenarios such as nursing homes.

[0029] 5. A feature extraction method based on multimodal technology is proposed to reduce computational burden while improving detection robustness;

[0030] In terms of vital sign signal feature extraction, after the data undergoes multi-level FFT, noise reduction filtering, and CFAR detection, this invention proposes a dual-modal branch structure based on different signal-to-noise ratio conditions:

[0031] (1) When the target thoracic cavity signal has a high signal-to-noise ratio, the variational mode decomposition method is used to significantly reduce the computational burden;

[0032] (2) When the target chest cavity signal has a low signal-to-noise ratio, the FPGA-based LSTM neural network method is used to mine hidden information, which effectively improves the robustness and accuracy of vital sign signal detection and is suitable for complex scenarios with multiple targets and dynamic changes.

[0033] 6. Achieve accurate tracking of multiple health statuses;

[0034] The host computer parses information packets in real time and displays the vital signs of multiple targets in a graphical interface. If a target's vital signs exceed an abnormal threshold or if there are abnormal movements such as falls, an alarm is issued to the administrator. The data management, analysis, and visualization functions implemented by the host computer provide nursing staff with an intuitive display of health status and abnormal alarms, improving the efficiency of health maintenance. Simultaneously, the long-term data recorded by the system provides data support for subsequent health management and disease early warning. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the monitoring system of the present invention;

[0036] Figure 2 This is a schematic diagram of the edge sensing node of the present invention;

[0037] Figure 3 This is a schematic diagram of the radar sensor module of the present invention;

[0038] Figure 4 This is a schematic diagram of the FPGA neural network inference module of the present invention;

[0039] Figure 5 This is a flowchart of the overall monitoring method of the present invention;

[0040] Figure 6 This is a flowchart illustrating the node task switching logic when the target of this invention is occluded.

[0041] Figure 7 This is a flowchart of the filtering algorithm for when vital signs are disturbed according to the present invention;

[0042] Figure 8 This is the real-time display interface of the host computer's visualization software in this invention;

[0043] Figure 9 This is the visualization software interface for displaying abnormal alarm records of the host computer in this invention.

[0044] In the figure, there are edge sensing node 1, host computer 2, microcontroller module 11, radar sensor module 12, DSP module 13, FPGA neural network inference module 14, communication module 15, and power supply module 16. Detailed Implementation

[0045] Specific embodiments of the present invention are given below. These specific embodiments are only used to further illustrate the present invention in detail and do not limit the scope of protection of the present invention.

[0046] The present invention provides a multi-unit radar collaborative motion state and vital signs monitoring system (hereinafter referred to as the system), characterized in that the system includes an edge sensing node 1 and a host computer 2 connected by communication;

[0047] Edge sensing node 1 has complete wireless sensing, signal processing and communication capabilities for vital signs. It is positioned at the same height as the human chest cavity (1.3 to 1.5 meters above the room to be monitored in this embodiment) to obtain the best radar signal strength. The host computer 2 is a personal computer that receives information such as target ID, target location, target vital signs and target abnormal status sent by edge sensing node 1.

[0048] The edge sensing node 1 consists of a microcontroller module 11, a radar sensor module 12, a DSP (digital signal processing) module 13, an FPGA neural network inference module 14, a communication module 15, and a power supply module 16. The microcontroller module 11 is communicatively connected to the radar sensor module 12, the DSP module 13, the FPGA neural network inference module 14, and the communication module 15. The radar sensor module 12 is communicatively connected to the DSP module 13. The DSP module 13 is communicatively connected to the FPGA neural network inference module 14. The power supply module 16 supplies power to the edge sensing node 1.

[0049] Preferably, the radar sensor module 12 continuously tracks and measures the minute displacement of the target's chest cavity surface, transmits frequency-modulated continuous waves and acquires raw data of reflected echoes in the scene, and the DSP module 13 calculates information such as the target position and target chest cavity displacement to extract heart rate and respiratory rate. The measurement results are input into the FPGA neural network inference module 14 for further feature extraction. The measurement results of all targets in the current scene acquired by the edge sensing node 1 are synchronized between nodes through the communication module 15, and the target ID, target position, target vital signs, target abnormal status and other information are sent to the host computer 2 through the communication module 15 via wired or wireless means through the root node for further statistics and display.

[0050] Preferably, the microcontroller module 11 is connected to the DSP module 13 via an SPI serial bus, to the FPGA neural network inference module 14 via the SPI protocol, and to the communication module 15 via the SDIO protocol; the DSP module 13 is connected to the radar sensor module 12 via a parallel interface.

[0051] Preferably, the microcontroller module 11 is composed of an STMicroelectronics high-performance microcontroller chip STM32H750XBH6 and its peripheral circuits. The microcontroller module 11 is the control core of the edge sensing node 1, realizing functions such as system initialization, data collection, data processing, and communication control. System initialization mainly completes the establishment of the communication link, sends radio frequency parameters and initialization commands to the radar sensor module 12, and activates the DSP module 13 and the FPGA neural network inference module 14. Data collection is used to control the sending of the raw data stream collected by the radar sensor module 12 to the DSP module 13 and the FPGA neural network inference module 14 for processing, and then reads the calculated personnel ID, vital signs, and abnormal status into the Flash memory. Data processing is used to format the data in the Flash memory and send it to the communication module 15.

[0052] Preferably, in step 1, the radar sensor module 12 includes a transmitting antenna, a power amplifier, a voltage-controlled oscillator, a phase shifter, a mixer, a linear amplifier, a receiving antenna, an A / D converter, and a microcontroller.

[0053] The voltage-controlled oscillator (VCO) is connected to both the power amplifier and the mixer. The transmitting antenna forms a transmitting link with the VCO via the power amplifier. The receiving antenna forms a receiving link with the mixer via a linear amplifier (a low-noise amplifier is used in this embodiment). The inputs of the two mixers are connected to the VCO and a phase shifter connected to the VCO, respectively. The outputs of the mixers are simultaneously connected to a quadrature demodulation circuit. The input of the A / D converter is connected to the quadrature demodulation circuit, and its output is connected to the parallel input interface of the DSP module 13. The radar uses frequency-modulated continuous wave (FM) with a working frequency of 60–64 GHz. Spatial information is acquired through a MIMO antenna array, and beamforming technology is used for multi-target thoracic cavity surface displacement measurement. The VCO transmits electromagnetic waves of a specified waveform according to the initialization command and RF parameters sent by the microcontroller module 11. After reflection on the target thoracic cavity surface, the electromagnetic waves are captured by the receiving antenna, down-mixed by the mixer to obtain the difference frequency component, and the I / Q signals are demodulated using arctangent to obtain the intermediate frequency signal. Then, after A / D conversion, the signal is sent to the parallel input interface of the DSP module 13.

[0054] Preferably, the DSP module 13 consists of a TI TMS320C6748 low-power DSP and its peripheral circuits. The DSP module 13 executes corresponding algorithms according to instructions sent by the controller module 11 and sends the processing results to the DMA channel of the controller module 11. The DSP module 13 is used to accelerate radar raw data signal processing algorithms, including range-dimensional FFT, Doppler-dimensional FFT, angle-dimensional FFT, beamforming, point cloud generation, target point cloud clustering, Kalman filtering, and variational mode decomposition algorithms.

[0055] Preferably, in step 1, the FPGA neural network inference module 14 consists of a Xilinx Zynq UltraScale+ field-programmable gate array and its peripheral circuits. The FPGA neural network inference module 14 is mainly used to deploy a feature extraction neural network model based on an LSTM model, find the features of the input vital signs signals and return respiratory rate, heart rate and fall detection, and send the output vital signs information, abnormal conditions and target ID back to the microcontroller module 11.

[0056] Preferably, the communication module 15 consists of an Espressif Systems WiFi module ESP32-S3-WROOM-1U-N16R8 and its peripheral circuitry. Upon receiving the initialization command from the microcontroller module 11, the communication module 15 activates its WiFi communication function and performs a Mesh self-organizing network, automatically scanning all available edge sensor nodes 1 within the signal range and forming a network. After successful network formation, all edge sensor nodes 1 can communicate with each other, exchange information, and achieve multi-node collaborative monitoring. When a target enters and causes a non-line-of-sight situation, the communication module 15 sends the relay status signal sent by the microcontroller module 11 to the edge sensor node 1 with the specified IP address for continuous monitoring. Furthermore, the node network will find the optimal path to transmit the monitoring data to the host computer 2.

[0057] Preferably, the power module 16 includes a lithium battery, a DC power supply circuit, and a voltage regulator circuit; it is powered by a power adapter and powered by the lithium battery when the power grid is unstable or there is a power outage, so as to realize all-weather monitoring of vital signs.

[0058] The host computer 2 is equipped with software; this software is a data visualization and management system written in C# and WPF, used to parse and statistically analyze information from edge sensor nodes 1, enabling data display and statistical functions. After running, host computer 2 connects to the gateway node or remote server node, establishing a TCP connection via a Socket interface to receive JSON-formatted data sent by all edge sensor nodes 1, and parses it into data such as target ID, target vital signs, target abnormal status, and target location. Host computer 2 includes a data transmission and parsing module, a node monitoring module, a data storage and retrieval module, a data statistics and display module, and an anomaly alarm module. Each module continuously operates after host computer 2 starts, implementing data display and statistical functions. The data receiving and parsing module is responsible for establishing a TCP connection with the root node via Socket, used to send control commands and receive JSON data sent by the node, and broadcast the parsed data to the interface; the node monitoring module is used to monitor the working status of each node; the data storage and retrieval module is used to save the received monitoring data according to the target ID, and can read the saved file to realize historical playback; the data statistics and display module is used to display the target ID, location, vital signs and abnormal status in the 3D interface, to intuitively count the current and historical status of the target for user analysis; the abnormal alarm module sends an alarm to the interface or sends a reminder to the user via SMS after receiving an abnormal value and abnormal status returned by a target.

[0059] This invention also provides a method for monitoring motion status and vital signs using multi-unit radar coordination (hereinafter referred to as the method), characterized by comprising the following steps:

[0060] Step 1: Set up the monitoring system;

[0061] Step 2: First, the radar sensor module 12 dynamically adjusts the measurement parameters according to the instructions issued by the microcontroller module 11, transmits signals and receives echo signals through frequency-modulated continuous wave; then, the radar sensor module 12 completes target information acquisition through mixing and sampling processing, and obtains a three-dimensional data matrix (in this embodiment, a RadarCube three-dimensional data matrix) composed of the original ADC data; subsequently, the DSP module 13 sequentially performs range dimension FFT (to calculate the distance), Doppler dimension FFT (to resolve velocity information), and angle dimension FFT combined with beamforming algorithm (to complete spatial orientation estimation) on the three-dimensional data matrix, and outputs a three-dimensional feature dataset containing target distance, angle and velocity;

[0062] Step 3: Input the three-dimensional feature dataset generated in Step 2 into DSP module 13 to perform multi-level preprocessing and output the primary target motion trajectory; then, based on the angle information in the primary target motion trajectory, perform beamforming based on the minimum variance distortionless response algorithm to enhance the target direction signal. On the basis of the original three-dimensional feature dataset, add the fourth dimension of the target chest cavity micro-motion feature to form a primary four-dimensional feature dataset containing the displacement information of the target chest cavity surface.

[0063] Preferably, in step 3, the multi-level preprocessing specifically involves: firstly, eliminating environmental interference through adaptive noise reduction filtering, and then using the CFAR constant false alarm rate detection algorithm to identify the coordinates of candidate targets; next, converting the triplet of distance, angle, and velocity of the candidate targets into three-dimensional spatial coordinates to generate an initial point cloud; then, using the DBSCAN density clustering algorithm to separate independent target point cloud clusters and establish a multi-target tracking queue; finally, using a Kalman prediction filter to perform temporal prediction of the target spatial coordinates, effectively suppressing measurement noise jitter and compensating for trajectory interruptions caused by short-term occlusion, and outputting the primary target motion trajectory.

[0064] Step 4: In order to build a multi-unit collaborative monitoring system to realize all-weather wide-area monitoring and analysis of human health status and abnormal status, the microcontroller module 11 performs multi-dimensional analysis on the primary four-dimensional feature dataset obtained in step 3 and establishes a target occlusion processing and monitoring relay mechanism; then, according to the target occlusion processing and monitoring relay mechanism, it processes the primary target motion trajectory and outputs a stable and continuous four-dimensional feature dataset.

[0065] Preferably, in step 4, the target occlusion handling and monitoring relay mechanism is as follows:

[0066] When the target's spatial coordinates have not been updated for 30 consecutive frames, the system calculates the target's motion vector based on the position sequence of the last 30 frames before the target disappears, and determines its positional relationship with the field of view boundary based on the following conditions:

[0067] When a target disappears within 5% of the field of view boundary, and the adjacent edge sensor node 1 detects a new target with similar motion characteristics to the disappeared target within 500 milliseconds of its field of view boundary, the microcontroller module 11 determines that the target has triggered a room switching behavior outside the line of sight. Subsequently, it triggers a monitoring relay mechanism, which synchronizes the target location information and vital signs information through communication between the edge sensor nodes 1, thereby achieving a seamless handover of the target tracking task, with the next edge sensor node 1 taking responsibility for monitoring the target.

[0068] When a target disappears into the field of view, or disappears into the 5% area of ​​the field of view boundary but does not appear in the field of view of the adjacent edge sensor node 1 within 500 milliseconds, the system marks it as a temporary occlusion state and initiates a hybrid prediction mechanism: In terms of target position prediction, the DSP module 13 calls the Kalman filter to perform kinematic extrapolation based on the target's historical velocity and acceleration, with a prediction time window of 1 second, to construct motion prediction data; in terms of vital sign prediction, the FPGA neural network inference module 14 analyzes the target's vital sign pattern, outputs the predicted value of the target's vital signs based on the target's historical vital sign data, and constructs vital sign prediction data; the motion prediction data and the vital sign prediction data together constitute the target prediction data group, which is used for future target ID rebinding;

[0069] When the target reappears within the field of view, the system triggers a target ID rebinding mechanism: the target prediction data set is input into the microcontroller module 11 for ID rebinding recognition. The microcontroller module 11 calculates the distance d between the reappearing target position and the Kalman filter predicted position and an adaptive threshold d. t The difference between the ratio of d to 1 t To verify the spatial prediction consistency of the target, a threshold distance calculated based on the target's historical maximum speed and the user-given tolerance is used. Simultaneously, the microcontroller module 11 constructs a vital sign parameter vector using historical and predicted data. This vector consists of heart rate, respiratory rate, heart rate variability, heart rate signal amplitude, and respiratory signal amplitude. Euclidean distance similarity is calculated between this vector and the reconstructed target's vital sign parameter vector to verify vital sign similarity. Then, the microcontroller module 11 performs a decision fusion mechanism based on these two similarity parameters. It obtains a weighted scoring model by multiplying the spatial prediction consistency parameter by the user weight α and the vital sign similarity parameter by the user weight β. When the score is greater than 0.85, the new target is confirmed as the original target, and the vital sign data continues to be updated; otherwise, it is marked as a new target, and a new ID file is created.

[0070] Step 5: Extraction of heartbeat and breathing features and fall detection features;

[0071] In the heartbeat and respiration feature extraction, firstly, based on the target chest cavity micro-motion features in the four-dimensional feature dataset output in step 4, variational mode decomposition is performed in DSP module 13 to obtain several intrinsic mode functions. In the decomposition results, if there are mode functions with energy concentrated in the 0.1-0.8Hz respiratory frequency band or the 1-3Hz heartbeat frequency band, and the frequency domain sidelobe suppression ratio is ≥20dB, they are determined to be valid vital sign signals; otherwise, they are determined to be low signal-to-noise ratio signals that do not meet the conditions, and are input to FPGA neural network inference module 14 for deep feature mining to obtain vital sign parameters. Then, target action influence correction is performed to obtain stable vital sign signals.

[0072] In the fall detection feature extraction, the four-dimensional feature dataset output in step 4 is differentially processed by DSP module 13 to extract the vertical velocity component of the human body's center of mass, and a two-layer detection mechanism based on kinematic criteria is constructed to determine whether a fall event has occurred.

[0073] Preferably, in step 5, variational mode decomposition specifically involves: extracting vital sign signal components with clear physical meaning by setting mode number constraints (3 to 4 intrinsic mode functions) and adaptive noise tolerance parameters (signal-to-noise ratio threshold set to 15dB).

[0074] Preferably, in step 5, the FPGA neural network inference module 14 is equipped with an LSTM neural network using a hierarchical feature extraction architecture; the LSTM neural network consists of a shallow spatiotemporal perception module, a deep feature coupling module, and a global feature fusion module;

[0075] This LSTM neural network model is jointly trained based on 500 sets of synchronously acquired millimeter-wave radar chest displacement data and measured data from a medical-grade chest strap sensor. The network establishes a nonlinear mapping relationship between the time-frequency domain features of the signal and physiological parameters, effectively extracting the heartbeat modulation features in the respiratory harmonic components.

[0076] The shallow spatiotemporal perception module consists of Bi-GRU (bidirectional gated recurrent unit) and ST-Conv (spatiotemporal convolutional layer). The Bi-GRU adopts a 64-unit structure and captures the short-term correlation of the signal through the tanh activation function and SiLU gating mechanism. The ST-Conv layer is configured with 5 groups of parallel time-frequency domain convolutional kernels and uses cross-channel max pooling to generate spatiotemporal joint features.

[0077] The deep feature coupling module consists of a two-layer LSTM and a phase synchronization unit. The first-layer LSTM uses a 128-element unidirectional LSTM to extract long-range time-dependent features of the radar signal, while the second-layer LSTM uses a 64-element bidirectional LSTM to capture bidirectional context information. Features are transferred between the two layers through skip connections. The phase synchronization unit directly analyzes the phase information of the radar signal. By calculating the instantaneous phase difference between adjacent time sequences, it dynamically fuses the phase difference with the hidden state of the second-layer bidirectional LSTM. The phase difference is used to weight and adjust the hidden state of the LSTM. When the radar signal phase changes abruptly, the heartbeat feature response of the LSTM at the corresponding moment is amplified by multiplication, and noise interference in the phase-stable interval is suppressed, thereby improving the detection accuracy of weak heartbeat signals in radar sensing.

[0078] The global feature fusion module employs a standard self-attention mechanism to dynamically weight the feature sequence after phase synchronization processing. By calculating the three-way attention weights of the time-frequency domain features and the hidden state, a nonlinear correlation mapping between features and heart rate and respiratory rate is established in the time-frequency dimension. This module uses a multi-head attention mechanism to focus on phase abrupt changes in the heartbeat harmonic frequency band, low-frequency oscillation modes of the respiratory envelope, and fundamental frequency stability features. Finally, the weighted and aggregated features are compressed into a 128-dimensional pulse-coded feature vector, where the first 64 dimensions encode the time-varying characteristics of the heartbeat rhythm, and the last 64 dimensions map the modulation mode of the respiratory waveform. In vital sign information... In the output stage, the 128-dimensional pulse-coded feature vector generated by the global feature fusion module is input to the dual-branch fully connected regression layer: the respiratory rate branch outputs the predicted respiratory rate value through a 64-32-1 fully connected network, and the heart rate branch outputs the predicted heart rate value using a bottleneck structure designed as 32-16-1; subsequently, the global feature fusion module synchronously receives the neural network prediction value, the Kalman filter prediction value, and the variational mode decomposition fundamental frequency parameter, and dynamically adjusts the fusion weights of the three based on the real-time signal quality index to generate vital sign parameters; the output result is smoothed by median filtering with a 1-second window length and then uploaded to the host computer 2 at a frame rate of 30Hz.

[0079] Preferably, in step 5, the specific steps of deep feature mining are as follows:

[0080] S51. Multimodal signal preprocessing: The target thoracic cavity surface displacement information of the primary four-dimensional feature dataset obtained in step 3 is used to generate 256 signal segments through a sliding window; synchronous time-frequency analysis is performed on each signal segment, and a time-frequency map is generated using complex Morlet wavelet transform (frequency resolution 0.1Hz, time window length 1 second), which is then concatenated with the first-order difference waveform of the original signal to form a two-dimensional feature matrix. After normalization, a 128×128×2 neural network input tensor is formed; the tensor is then input into the shallow spatiotemporal perception module to generate spatiotemporal joint features;

[0081] S52, LSTM Feature Extraction: The spatiotemporal joint features are input into the deep feature coupling module. First, the long-range temporal dependency features of the radar signal are extracted through the first layer LSTM. The bidirectional context is captured by the second layer LSTM and fused with the first layer through skip connections. Then, the phase change intervals caused by the heartbeat signal within the respiratory cycle are distinguished. In the phase change interval, the heartbeat feature response is amplified by multiplication, and noise is suppressed in the non-phase change interval to obtain the phase-modulated feature sequence.

[0082] S53. Cross-modal feature fusion and calibration: A global feature fusion module is used to dynamically weight the phase-modulated feature sequence to generate a 128-dimensional pulse-coded feature vector; then, a dual-branch regression output is used to input the pulse-coded feature vector into a fully connected regression layer to output heart rate and respiratory rate parameter values; then, the neural network prediction value, Kalman filter prediction value and variational mode decomposition fundamental frequency parameter are dynamically fused, and the weights of the three are adjusted according to the real-time signal quality index to generate vital sign parameters.

[0083] Preferably, in step 5, the correction method for the impact on the target action is as follows:

[0084] For measurement distortions caused by target actions (such as limbs obscuring the chest cavity, rapid turning, etc.), the correction method achieves suppression of the impact of target actions through multi-dimensional feature joint analysis: first, point cloud deformation analysis is used to identify abnormal motion frames and trigger an abnormal motion frame compensation mechanism. When the action interference ends, the system executes an iterative calibration method to finally obtain stable vital sign signals.

[0085] The method for identifying abnormal motion frames is as follows: when the acceleration value of a target point at the chest cavity in a certain frame exceeds 6 m / s², 2 When this occurs, it is determined to be a motion interference event and marked as an abnormal motion frame;

[0086] The abnormal motion frame compensation mechanism is as follows: During the abnormal motion frame, the DSP module 13 constructs a Kalman filter state space model (the state variables include heart rate, respiratory rate and their first derivatives) based on the effective vital sign data of the previous 30 seconds, dynamically adjusts the prediction weights through the time-varying process noise covariance matrix, and maintains the vital sign parameters according to the autoregressive prediction results of historical data; and during the abnormal motion frame, the physiological parameter fluctuation threshold of this period is automatically relaxed to avoid false alarms.

[0087] The iterative calibration method after the abnormal motion frame ends is as follows: First, calculate the residual sequence between the Kalman prediction value and the actual measurement value during the interference period. Update the observation noise matrix of the Kalman filter through backpropagation. Then, use a 15-second sliding window baseline to Gaussian weighted smooth the current vital sign parameters and force the outlier values ​​generated by the motion interference to regress to the baseline median.

[0088] Preferably, in step 5, the dual-layer detection mechanism specifically involves: first, calculating the average vertical velocity of 5 consecutive frames; when it exceeds 9.8 m / s... 2 Initial screening is triggered when the theoretical free-fall acceleration reaches 65%; then trajectory verification is performed by combining the height change rate (window length 0.5 seconds). If the height decrease exceeds 40% of the body height within 3 seconds and the velocity curve exhibits a parabolic characteristic, it is determined to be a fall event.

[0089] Step 6: Transmit the vital signs signals and whether a fall event has occurred obtained in Step 5 to the host computer 2 for display and statistics.

[0090] Any aspects not covered in this invention are applicable to existing technologies.

Claims

1. A method for monitoring motion status and vital signs using multi-unit radar coordination, characterized in that, The method includes the following steps: Step 1: Build a multi-unit radar collaborative motion status and vital signs monitoring system; the system includes an edge sensing node (1) with communication connection and a host computer (2). The edge sensing node (1) has complete wireless sensing, signal processing and communication capabilities for vital signs; the host computer (2) receives the target ID, target location, target vital signs and target abnormal status sent by the edge sensing node (1); The edge sensing node (1) is composed of a microcontroller module (11), a radar sensor module (12), a DSP module (13), an FPGA neural network inference module (14), a communication module (15), and a power supply module (16). The microcontroller module (11) is connected to the radar sensor module (12), the DSP module (13), the FPGA neural network inference module (14), and the communication module (15) respectively. The radar sensor module (12) is connected to the DSP module (13). The DSP module (13) is connected to the FPGA neural network inference module (14). The power supply module (16) supplies power to the edge sensing node (1). Step 2: First, the radar sensor module (12) dynamically adjusts the measurement parameters according to the instructions issued by the microcontroller module (11), transmits signals and receives echo signals; then the radar sensor module (12) completes the target information acquisition and obtains a three-dimensional data matrix composed of the original ADC data; then the DSP module (13) outputs a three-dimensional feature dataset containing the target distance, angle and velocity. Step 3: Input the three-dimensional feature dataset generated in step 2 into the DSP module (13) to perform multi-level preprocessing and output the primary target motion trajectory; then, based on the angle information in the primary target motion trajectory, perform beamforming based on the minimum variance distortionless response algorithm to enhance the target direction signal and form a primary four-dimensional feature dataset containing the displacement information of the target thoracic cavity surface. Step 4: The microcontroller module (11) performs multi-dimensional analysis on the primary four-dimensional feature dataset obtained in Step 3 to establish a target occlusion processing and monitoring relay mechanism; then, it processes the primary target motion trajectory according to the target occlusion processing and monitoring relay mechanism to output a stable and continuous four-dimensional feature dataset. Step 5: Extraction of heartbeat and breathing features and fall detection features; In the heartbeat and breathing feature extraction, firstly, based on the target chest cavity micro-movement features in the four-dimensional feature dataset output in step 4, variational mode decomposition is performed in the DSP module (13) to obtain several intrinsic mode functions; in the decomposition results, if there are mode functions with energy concentrated in the breathing frequency band of 0.1~0.8Hz or the heartbeat frequency band of 1~3Hz, and the frequency domain sidelobe suppression ratio ≥20dB, they are determined to be valid vital sign signals; otherwise, they are determined to be low signal-to-noise ratio signals that do not meet the conditions, and are input into the FPGA neural network inference module (14) for deep feature mining to obtain vital sign parameters, and then the target action influence correction is performed to obtain stable vital sign signals; In the fall detection feature extraction, the four-dimensional feature dataset output in step 4 is differentially processed by the DSP module (13) to extract the vertical velocity component of the human body's center of mass, and a two-layer detection mechanism based on kinematic criteria is constructed to determine whether a fall event has occurred. Step 6: Transmit the vital signs signal and whether a fall event has occurred obtained in Step 5 to the host computer (2) for display and statistics.

2. The method for monitoring motion status and vital signs using multi-unit radar coordination according to claim 1, characterized in that, In step 2, the beamforming algorithm is performed sequentially on the three-dimensional data matrix using distance dimension FFT, Doppler dimension FFT, and angle dimension FFT.

3. The method for monitoring motion status and vital signs using multi-unit radar coordination according to claim 1, characterized in that, In step 3, the multi-level preprocessing is as follows: First, environmental interference is eliminated by adaptive noise reduction filtering, and the CFAR constant false alarm rate detection algorithm is used to identify the coordinates of candidate targets; then, the triplet of distance, angle, and velocity of the candidate targets is converted into three-dimensional spatial coordinates to generate an initial point cloud; then, the DBSCAN density clustering algorithm is used to separate independent target point cloud clusters and establish a multi-target tracking queue; finally, a Kalman prediction filter is used to perform temporal prediction of the target spatial coordinates, effectively suppressing measurement noise jitter and compensating for trajectory interruptions caused by short-term occlusion, and outputting the primary target motion trajectory.

4. The method for monitoring motion status and vital signs using multi-unit radar coordination according to claim 1, characterized in that, In step 4, the target occlusion handling and monitoring relay mechanism is as follows: When the target's spatial coordinates have not been updated for 30 consecutive frames, the system calculates the target's motion vector based on the position sequence of the last 30 frames before the target disappears, and determines its positional relationship with the field of view boundary based on the following conditions: When the target disappears within 5% of the field of view boundary, and the adjacent edge sensor node (1) detects a new target with similar motion characteristics to the disappeared target within 500 milliseconds, the microcontroller module (11) determines that the target has triggered a room switching behavior outside the line of sight, and then triggers a monitoring relay mechanism. Through communication between the edge sensor nodes (1), the target location information and vital signs information are synchronized, thereby achieving seamless target tracking task handover, with the next edge sensor node (1) responsible for monitoring the target. When a target disappears into the field of view, or disappears into the 5% area of ​​the field of view boundary but does not appear in the field of view of the adjacent edge sensing node (1) within 500 milliseconds, the system marks it as a temporary occlusion state and starts a hybrid prediction mechanism: In terms of target position prediction, the DSP module (13) calls the Kalman filter to perform kinematic extrapolation based on the target's historical velocity and acceleration, with a prediction time window of 1 second, to construct motion prediction data; In terms of vital sign prediction, the FPGA neural network inference module (14) analyzes the target's vital sign pattern, outputs the predicted value of the target's vital signs based on the target's historical vital sign data, and constructs vital sign prediction data; The motion prediction data and the vital sign prediction data together constitute the target prediction data group, which is used for future target ID rebinding; When the target reappears within the field of view, the system triggers the target ID rebinding mechanism: the target prediction data set is input into the microcontroller module (11) for ID rebinding recognition, and the microcontroller module (11) calculates the distance between the reappearing target position and the Kalman filter predicted position. d With adaptive threshold d t The difference between the ratio and 1, d t To verify the spatial prediction consistency of the target, a threshold distance calculated based on the target's historical maximum speed and the user-given tolerance is used. Simultaneously, the microcontroller module (11) constructs a vital sign parameter vector using historical vital sign data and prediction data. This vector consists of heart rate, respiratory rate, heart rate variability, heartbeat signal amplitude, and respiratory signal amplitude. The Euclidean distance similarity is calculated between this vector and the reconstructed target's vital sign parameter vector to verify the vital sign similarity. Then, the microcontroller module (11) performs a decision fusion mechanism based on the spatial prediction consistency and vital sign similarity parameters. By multiplying the spatial prediction consistency parameter by the user weight α and the vital sign similarity parameter by the user weight β, a weighted scoring model is obtained. When the score is greater than 0.85, the new target is confirmed as the original target, and the vital sign data is updated. Otherwise, it is marked as a new target, and a new ID file is created.

5. The method for monitoring motion status and vital signs using multi-unit radar coordination according to claim 1, characterized in that, In step 5, variational mode decomposition specifically involves extracting vital sign signal components with clear physical meaning by setting mode number constraints and adaptive noise tolerance parameters.

6. The method for monitoring motion status and vital signs using multi-unit radar coordination according to claim 1, characterized in that, In step 5, the FPGA neural network inference module (14) is equipped with an LSTM neural network with a hierarchical feature extraction architecture; the LSTM neural network consists of a shallow spatiotemporal perception module, a deep feature coupling module and a global feature fusion module; The shallow spatiotemporal sensing module consists of Bi-GRU and ST-Conv. The Bi-GRU adopts a 64-unit structure and captures the short-term correlation of the signal through the tanh activation function and SiLU gating mechanism. The ST-Conv layer is configured with 5 groups of parallel time-frequency domain convolution kernels and uses cross-channel max pooling to generate spatiotemporal joint features. The deep feature coupling module consists of a two-layer LSTM and a phase synchronization unit. The first layer LSTM uses a 128-unit unidirectional LSTM to extract long-range time-dependent features of the radar signal, and the second layer LSTM uses a 64-unit bidirectional LSTM to capture bidirectional context information. Features are transferred between the two layers through a skip connection. The phase synchronization unit directly analyzes the phase information of the radar signal. By calculating the instantaneous phase difference between adjacent time sequences, it dynamically fuses the phase difference with the hidden state of the second-layer bidirectional LSTM. The phase difference is used to weight and adjust the hidden state of the LSTM. When the radar signal phase changes abruptly, the heartbeat characteristic response of the LSTM at the corresponding moment is amplified by multiplication, and noise interference in the phase-stable interval is suppressed, thereby improving the detection accuracy of weak heartbeat signals in radar sensing. The global feature fusion module uses a standard self-attention mechanism to dynamically weight the feature sequence after phase synchronization processing. By calculating the three-way attention weights of the time-frequency domain features and the hidden state, a nonlinear correlation mapping between features and heart rate and respiratory rate is established in the time-frequency dimension. This module uses a multi-head attention mechanism to focus on the phase change points of the heartbeat harmonic frequency band, the low-frequency oscillation mode of the respiratory envelope, and the fundamental frequency stability features. Finally, the weighted and aggregated features are compressed into a 128-dimensional pulse-coded feature vector, in which the first 64 dimensions encode the time-varying characteristics of the heartbeat rhythm, and the last 64 dimensions map the modulation mode of the respiratory waveform. In the vital signs information output stage, the 128-dimensional pulse-coded feature vector generated by the global feature fusion module is input to a two-branch fully connected regression layer: the respiratory rate branch outputs the predicted respiratory rate value through a 64-32-1 structured fully connected network, and the heart rate branch outputs the predicted heart rate value using a bottleneck structure designed with a 32-16-1 structure. Subsequently, the global feature fusion module synchronously receives the neural network prediction value, the Kalman filter prediction value and the variational mode decomposition fundamental frequency parameter, and dynamically adjusts the fusion weight of the three based on the real-time signal quality index to generate vital sign parameters; the output result is smoothed by median filtering with a 1-second window length and then uploaded to the host computer at a frame rate of 30Hz (2).

7. The method for monitoring motion status and vital signs using multi-unit radar coordination according to claim 6, characterized in that, In step 5, the specific steps of deep feature mining are as follows: S51. Multimodal signal preprocessing: The target thoracic cavity surface displacement information of the primary four-dimensional feature dataset obtained in step 3 is used to generate 256 signal segments through a sliding window; synchronous time-frequency analysis is performed on each signal segment, and a time-frequency diagram is generated by using complex Morlet wavelet transform, and then concatenated with the first-order difference waveform of the original signal to form a two-dimensional feature matrix. After normalization, a 128×128×2 neural network input tensor is formed; then the tensor is input into the shallow spatiotemporal perception module to generate spatiotemporal joint features; S52, LSTM Feature Extraction: The spatiotemporal joint features are input into the deep feature coupling module. First, the long-range temporal dependency features of the radar signal are extracted through the first layer LSTM. The bidirectional context is captured by the second layer LSTM and fused with the first layer through skip connections. Then, the phase change intervals caused by the heartbeat signal within the respiratory cycle are distinguished. In the phase change interval, the heartbeat feature response is amplified by multiplication, and noise is suppressed in the non-phase change interval to obtain the phase-modulated feature sequence. S53. Cross-modal feature fusion and calibration: A global feature fusion module is used to dynamically weight the phase-modulated feature sequence to generate a 128-dimensional pulse-coded feature vector; then, a dual-branch regression output is used to input the pulse-coded feature vector into a fully connected regression layer to output heart rate and respiratory rate parameter values; then, the neural network prediction value, Kalman filter prediction value and variational mode decomposition fundamental frequency parameter are dynamically fused, and the weights of the three are adjusted according to the real-time signal quality index to generate vital sign parameters.

8. The method for monitoring motion status and vital signs using multi-unit radar coordination according to claim 1, characterized in that, In step 5, the correction method for the impact of the target's actions is as follows: First, point cloud deformation analysis is used to identify abnormal motion frames and trigger an abnormal motion frame compensation mechanism. When the motion interference ends, the system executes an iterative calibration method to finally obtain stable vital sign signals. The method for identifying abnormal motion frames is as follows: when the acceleration value of a target point at the chest cavity in a certain frame exceeds 6 m / s², 2 When this occurs, it is determined to be a motion interference event and marked as an abnormal motion frame; The abnormal motion frame compensation mechanism is as follows: During the abnormal motion frame, the DSP module (13) constructs a Kalman filter state space model based on the effective vital sign data of the first 30 seconds, dynamically adjusts the prediction weight through the time-varying process noise covariance matrix, and maintains the vital sign parameters according to the autoregressive prediction results of historical data; and during the abnormal motion frame, the physiological parameter fluctuation threshold during this period is automatically relaxed to avoid false alarms. The iterative calibration method after the abnormal motion frame ends is as follows: First, calculate the residual sequence between the Kalman prediction value and the actual measurement value during the interference period. Update the observation noise matrix of the Kalman filter through backpropagation. Then, use a 15-second sliding window baseline to Gaussian weighted smooth the current vital sign parameters and force the outlier values ​​generated by the motion interference to regress to the baseline median.

9. The method for monitoring motion status and vital signs using multi-unit radar coordination according to claim 1, characterized in that, In step 5, the dual-layer detection mechanism specifically involves: first, calculating the average vertical velocity over five consecutive frames; when it exceeds 9.8 m / s... 2 Initial screening is triggered when the theoretical free-fall acceleration reaches 65%; then trajectory verification is performed in conjunction with the rate of change of height. If the height decreases by more than 40% of the body height within 3 seconds and the velocity curve exhibits a parabolic characteristic, it is determined to be a fall event.

Images