Heartbeat Interval Recognition via Radar Monitoring and Customized Wavelet Transform Template

Heart rate variability is a key indicator for assessing heart health. Bio-radar provides a solution for non-contact monitoring, and the main problem faced is how to achieve accurate inter-beat interval (IBI) estimation under low signal-to-noise ratio (SNR) conditions. Wavelet transform (WT) is an effective tool for analyzing non-stationary signals. However, the mother wavelet of traditional wavelet transform is generic and there is still room for performance improvement for the specific scenario of heartbeat detection. In this paper, we propose a method to customize the wavelet transform templates, and six mother wavelets are extracted from different individual real-world data. And the cross-correlation strategy is utilized to extract each heartbeat moment. It is shown experimentally that the root-mean-square error of IBI using the proposed method is on average 15.7 ms lower than that using Morlet wavelets when the subjects are sitting still and breathing normally.