Non-contact real-time heart rate estimation based on multi-channel FMCW radar sensor

Due to its good privacy protection feature, millimeter-wave radar sensor has an exciting prospect in the field of the noncontact heart rate (HR) monitoring. However, noisy environments add a tier of diffculty to HR estimation, so that it has to be carefully handled. To reduce noise interference, this paper exploits the correlation of the multichannel signals to enhance the HR estimation performance. We employ the multivariate variational mode decomposition (MVMD) algorithm that can decompose the multi-channel signals jointly to share the similar center frequency and bandwidth with all channels, which reduces the effect of random noise between channels in the signal decomposition phase. In the heartbeat signal reconstruction phase, we reconstruct the heartbeat signals of each channel by the energy ratio weighting, and then multiply them point by point to combine the final heartbeat signal. The random error of the heartbeat estimation of a single channel is eliminated by further considering the correlation between channels. Experimental results show that the proposed method effectively eliminates spurious peaks and the random error of the single-channel signal. Compared to several other methods, the proposed method achieves higher accuracy, with an average real-time HR detection error of less than 3 bpm.