Radar-based Breathing Signal Detection In Motion

Respiration is one of the important indicators that reflect physical conditions.Radar-based respiration detection does not require any equipment to be worn,and is flexible in application.It has become a popular detection method.When performing breathing detection in reality,the presence of body movement will cause huge interference.The existing radar breathing detection methods require the human body to be in a state of slight body movement such as sitting or lying still,which restricts the body’s posture.Aiming at this problem,this paper studies the removal of body movement under the motion state,and realizes the radar breathing detection under the motion state.The main work of this paper is as follows:1.Introduced the breathing detection system based on continuous wave radar,analyzed the working principle of the radar and the echo signal,analyzed the body movement removal algorithm based on matched filtering,LMS and wavelet transform in the static state,and the performance of the algorithm is compared and verified through simulation.Finally,the signal processing flow is given based on the radar breathing detection under the motion state targeted in this article.2.Aiming at the influence of big body movement(Doppler frequency shift generated by human motion)on the respiratory signal under the motion state,a big body movement removal algorithm based on adaptive frequency shift detection is designed,using a variable step sliding window method,the signal is adaptively divided according to the different stages and directions of human motion,and the Doppler frequency shift cancellation processing is carried out according to different movement states.Finally,simulations verify that the algorithm can effectively remove the influence of big body movement on the respiratory signal.3.Aiming at the influence of small body movement(random body movement)on the respiratory signal under the motion state,a small body movement removal algorithm based on deep neural network is proposed.First,the respiratory simulation signal used to establish the data set is optimized,and the data set establishment plan is designed,and then the DNN model is constructed,and the network structure and training parameters are tuned through the data set.4.Construct the experimental platform to verify the performance of the big body movement removal algorithm based on adaptive frequency shift detection,the performance of the small body movement removal algorithm based on deep neural network,and the feasibility of radar breathing detection under the motion state.The actual human body measurement results show that after the signal is processed by the big body movement removal algorithm proposed in this paper,the error between the obtained respiratory signal frequency and the frequency measured by the breathing belt is 3.0%;After processing the signal with the small body movement removal algorithm proposed in this article,the average RMSE between the obtained breathing signal and the signal measured by the breathing belt is reduced by 8.8% compared with the traditional random body movement removal algorithm,reaching 0.307;After using the method proposed in this paper to process the signal detected by the radar in the motion state,the error between the obtained respiratory signal frequency and the reference signal frequency measured by the breathing belt is 6.7%.