부산대학교 도서관

In recent years, we have seen efforts made to monitor respiration for multiple users. Existing approaches capture chest movement relying on signals directly reflected from chest or separate breath waves based on breath rate difference between subjects. However, several limitations exist: 1) they may fail when subjects face away from the transceiver or are blocked by obstacles or other subjects; 2) they may fail to separate subjects’ breath waves with the same or similar rates (i.e., breath rate difference < 1 bpm); 3) they assume a priori knowledge of number of subjects and cannot adapt to dynamic change of subject number during monitoring. To overcome these limitations, in this paper we propose MultiResp , a multi-user respiration monitoring system using acoustic signal. By fully leveraging the abundant acoustic signals reflected indirectly from subjects’ chest, MultiResp can robustly capture chest movement even when they face away from the transceiver or are blocked. By extracting fine-grained breath rate and phase difference between different subjects, MultiResp can separate the breath waves with the same or similar rates and adapt to dynamic change of subject number during monitoring. Extensive experiments show that MultiResp is able to accurately monitor the respiration of multiple users with a median error of 0.3 bpm in various indoor scenarios, however, it fails when the sound pressure is lower than 55 dB or body movement is happening.