BreathID: Respiration Sensing for Human Identification Using Commodity Wi-Fi

Human identification is essential for current intelligent applications. Traditional human identification methods usually require dedicated collection equipment or user cooperation, which limits the application scope of human identification. Recent research works have shown that Wi-Fi signals can support fine-grained sensing for human respiration. In this article, we leverage Wi-Fi signals to collect and recognize a unique physiological feature of humans, i.e., the respiration, for device-free and nonintrusive human identification. We propose a respiration sensing-based human identification system, namely BreathID. BreathID explores the nature of the amplitude and phase data in the CSI ratio model through vector analysis. By jointly using the discrete wavelet transform (DWT) and fake peak removal method, we accurately calculate the respiration rate and extract the breathing feature. We also propose a weighted multidimensional dynamic time warping (WMD-DTW) algorithm to realize the human identification based on multiple input multiple output (MIMO) technology. We implement the BreathID system on commercial Wi-Fi devices. The results show that BreathID can identify 11 users in three typical scenarios in daily life with an average accuracy of 97.52%. In particular, the average absolute error of the respiration rate in BreathID is about 0.1 bpm, which is better than the state-of-the-art algorithms.