Research On Life Detection Technology Based On Ultra-Wide-Band Radar In Sheltered Environment

In the post-disaster rescue scene and medical health monitoring,due to the limitation of environmental conditions and the particularity of the human target,it is often necessary to detect the life information of the human target through shelters such as ruins,rubble,and walls.Ultra-wide band radar is widely used in life detection technology in sheltered environments because of its high range resolution,strong penetrability,and strong anti-interference.Based on ultra-wide band radar,this thesis conducts research on human life detection technology in sheltered environments.The main research contents are summarized as follows:1.The experimental platform of the ultra-wide band radar life detection system was built,and the wall was selected as the shelter,and the radar echo data of the human target in different directions and different breathing states were collected,and the breathing sample data set was established.Aiming at the problem of low signal-to-noise ratio of echo signal in ultra-wide band radar life detection technology in sheltered environment,an echo signal preprocessing method based on S-G filter-singular value decomposition(S-G-SVD)is proposed.Firstly,This method performed traditional clutter suppression on the echo signal,and used singular value decomposition(SVD)to decompose the signal after clutter suppression to obtain principal components of each order of echo signal,and calculated the energy of principal components of each order,according to the energy weighted reconstruction of the signal;secondly,savitzky-golay filter is introduced into the reconstructed signal,and polynomial fitting method is used to smooth and denoise the reconstructed signal;finally,the S-G-SVD algorithm is used to process the simulation modeling data and the data collected in the experiment.By calculating the signal-to-noise ratio of the echo signal before and after preprocessing,the results showed that the signal-to-noise ratio of the echo signal processed by S-G-SVD has been greatly improved.2.Aiming at the problem that the weak vital signs are difficult to be accurately extracted in an occluded environment,a vital sign extraction method based on improved time-varying filtering empirical mode decomposition(ITVFEMD)is proposed.This method aimed at the parameter uncertainty problem in the time-varying filtering empirical mode decomposition(TVFEMD)method,firstly,the energy percentage of the human respiration and heartbeat signal is used as the fitness function;secondly,according to the principle of the TVFEMD algorithm and the characteristics of human vital sign signals,the fitness function is converted into an objective function related to the parameters(bandwidth threshold and B-spline order)in TVFEMD;finally,the parameter optimization is carried out by calculating the maximum value of the objective function.Under the optimal parameter combination after optimization,TVFEMD decomposition is performed on the maximum range gate signal.In order to eliminate the influence of respiratory harmonics on the heartbeat signal,the energy of the intrinsic mode function(IMF)component obtained by ITVFEMD decomposition is calculated in the frequency domain,and the energy proportion is used as the weight coefficient,and the appropriate IMF component is selected for weighted reconstruction of breathing and heartbeat signals,the method is verified by theoretical modeling data and experimentally collected radar echo data.The experimental results proved that the method can extract breathing and heartbeat signals in the sheltered environment.The consistency analysis of the data actually measured by the oximeter showed that the average estimation accuracy of the breathing rate and heart rate has all reached more than 93%.3.Aiming at the problem of recognition of breathing state of human targets in sheltered environments,firstly,the time-domain characteristics of radar echo signals are introduced;secondly,combined with the advantages of neural networks,using self attention convolutional neural network with bidirectional long and short term memory network based on the time-domain characteristics of the signal(SA-CNN-BiLSTM)classifies the collected echo signals;finally,for the three breathing states of the human target,such as fast respiration,normal respiration and slow respiration,the average recognition accuracy reached 96.00%,which provides a theoretical basis for the research on the recognition of breathing state based on ultra-wide band radar in sheltered environments.