Multi-human Target Vital Sign Detection Technology Of Millimeter Wave Radar

Respiration rate and heart rate,as important physiological assessments of the humans,providing valid a priori knowledge for the diagnosis of many diseases.However,most of the current research has focused on the measurement of single human target vital signs,and the detection of multiple human target vital signs has been comparatively weakly studied.In this thesis,a non-contact multi-target vital sign detection technique using Frequency Modulated Continuous Wave(FMCW)radar is proposed,and the main research elements are as follows:For the interferences between multiple human targets,the problem of not being able to accurately detect and separate different human targets.Three-dimensional fast Fourier transform(3D-FFT)method is proposed to estimate the number and angle of targets.At the same time,Linear Constrained Minimum Variance-based Adaptive Beamforming(LCMVADBF)method is used to weaken the inter-target interference and enhance the human target signal,while Constant False Alarm Rate(CFAR)method is used to detect and separate the targets.The problem that the breathing and heartbeat signals are disturbed by their own jitter and noise after target separation is addressed.The chest position of the human body is detected by fast Fourier transform,and the phase of the vital sign signal measured at the chest position is corrected using a circle tracking algorithm.Subsequently,the extended differential and cross-multiply(DACM)algorithm is used to demodulate the phase,avoiding the DC offset and phase ambiguity problems.Based on this,a singular spectrum analysis method based on eigenvalue selection is proposed to effectively suppress the interference and noise in the vital sign signals.To address the difficulty of separating respiratory and heartbeat signals.An improved complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)method is proposed to achieve the vital sign signal reconstruction as well as to calculate the respiration and heartbeat frequencies by the autocorrelation method.The data collected by radar are processed offline,and the following experimental results are obtained.For single human target detection,the levels of agreement of the respiratory and heartbeat compared with the reference results are 98.1% and 97.4%,respectively.With two human targets,the levels of agreement of the respiratory and heartbeat obtained by the proposed method and the reference results are 92.7% and 91.2%,respectively.When three human targets are present,the levels of agreement between the respiratory and heartbeat obtained by the proposed method and the reference results are 89.6% and 85.3%,respectively.In addition,in order to verify the effect of this scheme in practical application,the collected single human target data are processed in real time,the levels of agreement between the real-time respiration and heartbeat rates and the reference results are 97.2% and 95.4%,which fully demonstrates the effectiveness of the proposed method.