Breath And Heart Detection And Gesture Recognition Based On Millimeter-wave Radar

Millimeter wave radar plays a great role in the industrial field due to its small size,high precision,all weather,all day and non-contact type.In escalating times,people are more concerned about physical health.In the case of non-contact,millimeter-wave radar can get human breathing and heart rate;Technology is advancing.Millimeter wave radar can also realize the function of gesture recognition without infringing on user privacy.Therefore,this paper mainly researches on the method of breath and heartbeat detection and gesture recognition based on millimeter wave radar.The research contents are as follows:1.Introduce the composition of the millimeter wave radar and the principles of ranging,speed and angle measurement.In the research of breathing and heartbeat detection methods,this paper studies the method of breathing and heartbeat waveform extraction.Aiming at the problem of phase winding phenomenon in the phase information extraction,this paper studies the method of phase unwinding to extract undistorted respiratory and heartbeat waveforms.Aiming at the problem of the short range of action in the radar breathing and heartbeat detection method,this paper demonstrates the long-distance respiration and heartbeat waveform extraction using coherent accumulation,which increases the range of action.2.Analyze the characteristics and shortcomings of the detection of respiratory and heartbeat.Aiming at the problem that breathing harmonics would drown the weak heartbeat signal,this paper proposes a breathing and heartbeat detection method based on the combination of notch and optimal filtering.This paper designs a breathing and heartbeat detection program to analyze and verify the measured data.3.In the research of gesture recognition methods,this paper studies the feature value extraction method of gesture distance,speed and angle and analyzes the principle of gesture recognition.Different factors such as gestures,positions and amplitudes need to be considered when constructing the data set.Based on the characteristics of measured data analysis,the problem of high data integration cost is solved by constructing simulation data.4.In the case of a small data set,the recognition method based on custom convolutional neural network and the gesture recognition method based on transfer learning are studied.The results show that the comprehensive accuracy of gesture recognition can reach 89.3%and 96.6%,respectively.In order to improve the accuracy of gesture recognition,this paper proposes a gesture recognition method that uses a simulation model to expand the gesture data set.The results show that the comprehensive accuracy rate of gesture recognition is99.1%.