Design and Implementation of a CSI-Based Ubiquitous Smoking Detection System

Xiaolong Zheng,Jiliang Wang,Longfei Shangguan,Zimu Zhou,Yunhao Liu

DOI: https://doi.org/10.1109/tnet.2017.2752367

2017-01-01

IEEE/ACM Transactions on Networking

Abstract:Even though indoor smoking ban is being put into practice in civilized countries, existing vision or sensor-based smoking detection methods cannot provide ubiquitous detection service. In this paper, we take the first attempt to build a ubiquitous passive smoking detection system, Smokey, which leverages the patterns smoking leaves on WiFi signal to identify the smoking activity even in the non-line-of-sight and through-wall environments. We study the behaviors of smokers and leverage the common features to recognize the series of motions during smoking, avoiding the target-dependent training set to achieve the high accuracy. We design a foreground detection-based motion acquisition method to extract the meaningful information from multiple noisy subcarriers even influenced by posture changes. Without the requirement of target's compliance, we leverage the rhythmical patterns of smoking to detect the smoking activities. We also leverage the diversity of multiple antennas to enhance the robustness of Smokey. Due to the convenience of integrating new antennas, Smokey is scalable in practice for ubiquitous smoking detection. We prototype Smokey with the commodity WiFi infrastructure and evaluate its performance in real environments. Experimental results show Smokey is accurate and robust in various scenarios.