Non-contact Heartbeat Detection Using Viterbi Algorithm Based on Distribution of Difference of Two-Adjacent R-R Intervals

Abstract

Vital sign monitoring is getting more and more attention in the field of health care. The variation of RRI (R-R Interval) is one of the vital signs that can represent mental stress conditions and heart diseases. Many non-contact Doppler sensor-based heartbeat detection methods have been proposed to evaluate the information of RRI without the device attachment. However, unwanted peaks due to respiration and small body motion could appear over the signal obtained by some signal processing, even when a subject keep still with normal breathing. In this paper, we propose a selection method of heartbeat peaks by the Viterbi algorithm. We confirmed that a Gaussian distribution could approximate the difference of two adjacent RRIs through the preliminary experiments. Based on this fact, as BM (Branch Metric) in the Viterbi algorithm, we use the squared difference of two adjacent RRIs. We evaluate our peak selection method in several peak detection methods such as (i) Spectrogram-based peak detection method and (ii) Doppler output after LPF (Low-Pass Filter)-based peak detection method. According to the experimental results, we show that our method, “Viterbi with squared BM” is effective for each peak detection method. We also show that “Spectrogram + Viterbi with squared BM” outperforms the “Doppler output after LPF + Viterbi with squared BM” method in terms of RMSE (Root-Mean-Square Error) of RRIs.