Research On Multi-person Vital Sign Detection Based On Blind Source Separation Technolog

Vital sign detection for multiple persons has a great demand in the fields of healthcare applications and health monitoring.Radar-based sensing has attracted a lot of interest in recent years because of its advantages of noncontact and high precision.However,limited by range resolution and angle resolution of radar system,monitored people must stay away from each other,which prevented applications in many scenes.This paper proposed a multi-person vital signs detection method based on blind source separation technology,which can achieve and separate respiration signals of multiple persons under the same range gate and beam.The main research contents are summarized as follows:1.Introduced single-transmit and multiple-receive FMCW radar system.Then the system structure and principle were explained.The radar echo signal was analyzed theoretically,and the preprocess of the radar echo signal was implemented through data rearrangement and phase extraction.2.Based on blind source separation model,the radar echo phase signal including vital signs of multiple persons was modeled.Phase signal was reconstructed by the maximum overlap discrete wavelet transform,so that the reconstructed signal can satisfy the constraints of the blind source separation model.Determined BSS algorithm was researched.Based on the second-order statistics of signal,the target tensor was constructed and tensor-based algorithm was implemented.The performance of ICA,JADE and tensor-based algorithm were compared and analyzed through simulation.The simulation results showed that the tensor-based algorithm had better performance.3.The multi-person vital signs detection was researched based on underdetermined BSS model.Combining the characteristics of the respiration signals and deficiencies of the sparse component analysis,an improved sparse component analysis algorithm was proposed.Firstly,the sparseness of the observation space was increased through single source point recognition;then an improved K-means algorithm was proposed to avoid the problem that clustering result was sensitive to outliers and initial cluster centers.Furthermore,the number of people and mixing matrix were estimated adaptively based on the improved affinity propagation algorithm.Finally,according to the estimated mixing matrix,the respiration signals of multiple people were restored under the constraint of the minimum L1 norm.4.An experimental platform for multi-person respiration signal detection based on SIMO radar system was built.In the determined BSS algorithm verification experiment,the respiration signal of two subjects were separated accurately using two receive antennas.The maximum measurement error of respiration frequency was 0.024 hz.The correlation coefficient between separated signal and reference signal was no less than 0.91.In the underdetermined BSS algorithm verification experiment,the respiration signal of three subjects were separated accurately using two receive antennas.The maximum measurement error of respiration frequency was 0.0186 hz.The correlation coefficient between separated respiration signal and reference signal was no less than 0.87.The experimental results verified the effectiveness of the radar system and the proposed algorithm in this paper.