Multi-person breathing rate estimation using time-reversal on WiFi platforms

In this paper, we present TR-BREATH, a time-reversal (TR) based, contact-free, accurate breathing monitoring system capable of multi-person breathing rate estimation within a short period of time (e.g., around one minute) using off-the-shelf WiFi devices. TR-BREATH exploits the fine-grained channel state information (CSI) on WiFi devices to capture the minor variations caused by breathing. To amplify such variations, TR-BREATH projects CSI time series into TR resonating strength (TRRS) feature space and performs Root-MUSIC to extract candidates of breathing rates. Then, TR-BREATH performs affinity propagation to partition these candidates into clusters corresponding to the breathing of different people. Extensive experiment results in a typical indoor environment demonstrate that under an non-line-of-sight scenario, TR-BREATH achieves an accuracy of 98.5% in estimating the breathing rates of one person, and a mean accuracy of 96.9% averaging over the accuracies of 1 to 7 people.