Wi-Tracker: Monitoring Breathing Airflow with Acoustic Signals

Abstract

Continuous and accurate breathing monitoring during sleep plays a critical role in early warning and health diagnosis of diseases. To extract breathing patterns, there exist three categories of solutions, i.e., using camera to collect image data, wearing sensor devices or deploying dedicated hardware to record sensor data. However, video-based schemes raise privacy concerns and sensor-based schemes are intrusive. Deploying dedicated hardware incurs high cost. In addition, most of the existing solutions focus on sensing chest/abdomen movement to detect breathing. However, chest/abdomen movement is not a good indicator for breathing monitoring due to existing false body movement. To overcome the above limitations, we propose Wi-Tracker, a contactless and nonintrusive breathing monitoring system, which exploits ultrasound signals generated by smartphone to sense Doppler effect caused by exhaled airflow on the reflected sound waves to detect breathing. By analyzing the data collected from real sleep environments, we find that Power Spectral Density (PSD) of acoustic signals can be utilized to sense breathing procedures. Specifically, Wi-Tracker first adopts a Cumulative PSD method to eliminate frequency interference and extract breathing patterns. Then, Wi-Tracker designs a CPSD-based peak detection algorithm to detect breathing events. Finally, Wi-Tracker applies a fake peak removal algorithm to further improve its performance. We evaluate Wi-Tracker with six volunteers over a one-month period. Extensive experiments show that Wi-Tracker can achieve a Mean Estimation Error (MEE) of 0.17 bpm for breathing rate estimation, which is comparable or even better as compared to existing WiFi-based or RF-based approaches.