Research On Non-contact Health Monitoring System Based On Millimeter-wave Radar

In recent years,the health monitoring of the elderly living alone has become a research hotspot.However,the current common solutions based on wearable devices or cameras have some shortcomings: wearable devices are affected by battery life and the habits of the elderly,so it is difficult to play an effective role at any time;it is difficult for cameras to be implemented in dark and highly private places.Millimeter-wave radar has attracted attention due to its advantages of strong privacy,but most of the current research in health monitoring is based on offline data and is conducted for a single problem.Given the above problems,this thesis designs a set of elderly health monitoring systems based on FMCW millimeter-wave radar,which realizes three functions of vital sign detection,personnel tracking,and fall detection.For this system,this thesis mainly does the following work from the aspects of algorithm research and design,experimental simulation,system framework design,and performance verification:First of all,in terms of vital signs detection,the VME algorithm is introduced to solve the problems that the short-time Fourier transform algorithm and several types of modal decomposition algorithms are interfered by the respiratory harmonics in the heartbeat signal,and the time complexity of the algorithm is too high and difficult to realize in real-time.The comprehensive performance of the algorithm in terms of accuracy and real-time performance is proved to be better than the above-mentioned algorithms through the measured data.Then,the millimeter-wave radar point cloud imaging algorithm is studied.To solve the problem of poor angular resolution of radar systems,TDM-MIMO technology is used to enhance the angular resolution of radar systems.And the 2D-FFT algorithm and 2DCFAR algorithm are used to obtain the distance and velocity information of the target.On this basis,the 3D-FFT algorithm is used to estimate the target's pitch angle and azimuth angle.Finally,richer point cloud information is obtained by the method of multi-frame stacking.Next,the tracking and fall detection research is carried out based on the radar point cloud data.In terms of point cloud clustering,to solve the problem that the different density of point cloud clusters affects the clustering effect,the OPTICS clustering algorithm is adopted after simulation and comparison.In the aspect of multi-target tracking,the tracker,track initiation,data association,and track management are studied and simulated.The fall detection part is based on the previous tracking results,obtains the three-dimensional centroid sequence,and models the radar fall detection problem as a time series classification problem,effectively reducing the model complexity and improving detection speed.Finally,a one-dimensional CNN model is designed to Implement fall detection.Finally,the above functions are implemented based on python,a real-time elderly health monitoring system is designed,and the system's running speed and function isolation are improved through multi-threading design.Then,the system's vital signs detection,positioning and tracking,and fall detection functions are verified and analyzed.The average accuracy of breathing is 96.3%,the average accuracy of heartbeat is 97.9%,the average positioning accuracy is 0.1m,and the average fall detection is successful.The rate is 95.5%.The above results all reach the design targets.