Heartbeat detection by using Doppler radar with wavelet transform based on scale factor learning

In this paper, we focus on the scale factor in the wavelet transform. When we observe the time series of wavelet coefficients on each scale factor, the wavelet coefficients are affected by the respiration or the body motion on some scale factors. On the other hand, on some scale factor, the wavelet coefficients increase only when the heartbeats appear. We search the scale factor whose wavelet coefficients increase only when the heartbeats appear, in advance. In the learning phase, the subject sits still without body motion. We avoid using a pseudo frequency to obtain the heart rate. Instead, we use the time interval of each peak on the wavelet coefficients whose scale factor is decided in the learning phase. Thus, although the subject has body motion, the heartbeats are able to be detected. For the evaluation, four types of activities are tested. The R-R interval is used for the evaluation of heartbeat detection. As a result, we confirm that RMSE of the R-R interval reduced on all activities compared to the conventional method. Moreover, the RMSE of the R-R interval did not deteriorate when the distance between the Doppler radar and the subject becomes long.