Research On Target Recognition Algorithm Of Life Detection Radar

Life detection radar has been widely used in disaster rescue,medical monitoring,military anti-terrorism and other fields in recent years,in which the radar echo is increasingly concerned about the extraction of living target life signal and target identification.However,the current research on life detection radar is mainly limited to target detection without obstacles.At the same time,the classification of living targets based on respiratory and heartbeat rate mainly depends on the visual detection of operators.In combat,anti-terrorism and rescue operations,it has important military and civil value to quickly obtain the breathing and heartbeat characteristics of life behind the wall or under the ruins and automatically identify human and animal targets.Therefore,this paper focuses on the feature extraction and identification of life target of through wall radar.Firstly,the mathematical model of UWB stepped frequency radar signal and the principle of high range resolution imaging are introduced.In addition,the mathematical model of breathing,heartbeat and body movement of life target is established.Based on this,the radar echo is established,and the analytical formula of radar echo signal is deduced,which provides a theoretical model basis for the extraction and classification of breathing and heartbeat features.In order to solve the problem that the signal intensity of respiration and heartbeat is obviously lower than that of the echo and noise of stationary obstacles under the condition of passing through the wall,this paper studies a life feature extraction algorithm combining feedback pulse cancellation,zero frequency clutter elimination,multiple autocorrelation de-noising,and empirical mode decomposition feature extraction based on the respiration and heartbeat model established in Chapter 2.The feasibility of the algorithm is verified by MATLAB simulation and experimental data.In order to solve the problem that the frequency of respiratory and heartbeat micro motion signal is very low and difficult to distinguish,this paper explores a new time-frequency analysis method synchronous squeeze S transform(SSST),which has achieved good results in power signal analysis and seismic signal processing.In this paper,SSST is applied to analyze the time-frequency distribution of respiration and heartbeat.By comparing the time-frequency analysis effect of SSST with STFT,it is found that SSST frequency resolution is obviously better than STFT,which is more suitable for low-frequency complex signal analysis.Finally,the time-frequency map classification algorithm based on deep learning is studied.In this paper,the convolution recurrent network(RCNN)used in character image recognition is improved to make it suitable for feature extraction and classification of time-frequency distribution map.Experimental results show that the recognition accuracy of the model is 96%,which is 5.65%and 4.25% higher than that of traditional SVM and convolutional neural network classification.As an example to support the above research work,this paper uses the stepped frequency radar with frequency range of 0.9-1.6Ghz,frequency points with 351 to collect a large number of experimental data,and extracts the measured characteristics of human and animal respiration and heartbeat for experimental classification test,which verifies the algorithm theory in this paper.