Heart Rate Variability Extraction using Commodity Wi-Fi Devices via Time Domain Signal Processing

Heart rate is one of the most important indicators of health status. In place of conventional contact-based devices, contactless vital sensing methods that utilize radio waves have been attracting attention. Among them, channel state information (CSI) available with commodity Wi-Fi devices can be used for measuring heart rate. However, existing methods fail to capture individual heartbeat intervals, though it is the most useful information for health and emotion recognition. The fast Fourier transform with a long time window has been necessary to overcome the inherently low signal-amplitude of the heartbeat in CSI. In this study, we propose a time domain method to estimate heart rate variability (HRV) using commodity Wi-Fi devices. The proposed method combines two techniques: (1) denoising by CSI quotient model, and (2) signal selection based on a standard deviation of NN intervals (SDNN) and averaging. Experiments show that the proposed method for the first time achieves accurate estimation of HRV of a 130ms average RMS error, demonstrating that commodity Wi-Fi devices can be used to extract HRV.