Human Vital Signs Detection Based On FMCW Millimeter Wave Radar In Complex Environment

The vital sign detection method based on radar mainly uses radar signals to get the micro-motion characteristics of the target.When the target is a human,the breathing signal and heartbeat signal of the human body can be obtained from the phase of the received signal.Although many researchers have provided lots of effective and innovative methods in this field,traditional methods can't detect human physiological characteristics well in some complex situations,such as the situation of low Signal-to-Noise Ratios(SNR)and the situation of multiple human targets,etc.Therefore,this thesis mainly conducts the following researches around these two situations.(1)Establishing the echo model of human vital signs based on FMCW radarAccording to the principle of radar target detection,this thesis analyzes the influence of human body's micro-motion characteristics on radar signals,and theoretically establishes the extraction process of human physiological signals.In order to ensure the reliability of the experimental model,the thesis also studies the noise distribution in the specific environment,so as to ensure that the noise distribution in the model is consistent with the environmental noise.Human body echo model,noise model and related configuration of radar equipment,etc.,provide instructive suggestions for follow-up research work.(2)Vital sign detection in Low SNRFor the problem of single target detection in the case of low SNR,this thesis analyzes some defects of traditional vital signs detection algorithm in the low SNR situation.On this basis,according to the principle of modal decomposition,this thesis proposes a vital sign detection algorithm based on symplectic geometric transformation.The algorithm gives some optimization methods to improve the SNR in several aspects,including the filter method and the spectral estimation method,and finally separates the breathing signal and the heartbeat signal through the modal decomposition method of symplectic geometric transformation.(3)Vital sign detection algorithm based on deep networkIn addition to the signal processing method,in order to solve the problem of low SNR,this thesis also uses a deep network to denoise the signal.In this process,a physiological signal denoising algorithm based on generative adversarial network is proposed.In this method,the time-frequency data of the radar signal with low SNR is input into the denoising network,then the time-frequency result with high SNR is generated,and the phase signal is recovered from the original time-frequency data by image processing.This method uses a deep network to denoise the signal,improves the SNR of the phase signal,and further improves the detection effect of single-target vital signs under the condition of low SNR.(4)Research on multi-target vital signs detection methodIn the case of multiple targets,this thesis proposes an adaptive fast angle estimation algorithm for the problem of how to separate multiple targets.The algorithm combines multiple signal classification(MUSIC)algorithm and angle FFT algorithm.When estimating the target unit,some redundant data in the radar signal is removed by means of information dimensionality reduction.On the premise of ensuring the accuracy of azimuth estimation,the operation speed of the algorithm is accelerated.