Heartbeat detection with Doppler radar based on spectrogram

Abstract

A variability of R-R intervals that represent the peak-to-peak intervals of the heartbeats indicates the mental condition. Doppler radar can capture the information of heartbeats with less burden on subjects, which leads to less stress of subjects. However, non-contact heartbeat detection using Doppler radar is easily affected by respiration and body movements. In this paper, we propose a detection algorithm of R-R intervals based on the spectrogram. Our algorithm determines the frequency bands containing the heartbeats components from the frequencies that might respond to heartbeats in the spectrogram. We integrate the amplitudes of frequencies due to heartbeats within the frequency band to eliminate the noise caused by respiration and small body movements. Then, we detect peaks in the integrated amplitudes of frequencies corresponding to heartbeats. In general, the R-R intervals do not largely change between two adjacent intervals. Thus, we set a threshold to the difference of two adjacent peak-to-peak intervals that are detected. If the peak-to-peak interval is judged not corresponding to an R-R interval by the threshold, we remove the corresponding peak and interpolate a peak based on the adjacent peak-to-peak intervals. Through experiments, we show that when the subjects were sitting still, our algorithm improved the detection accuracy of the R-R intervals compared with our previous algorithm that was able to achieve a better detection accuracy than the other existing algorithms. Moreover, we confirmed that the improvement of the detection accuracy is effective to accurately calculate the stress index.