Wi-Run: Multi-Runner Step Estimation Using Commodity Wi-Fi

Abstract

Step counting is a fundamental unit of human locomotion, and is a preferred metric for quantifying physical activity. However, the existing step counters are too inconvenient to wear and the treadmill can not count the steps. Recently, commercial Wi-Fi based device-free sensing shows a promising future for ubiquitous motion-based interactions and provides possibility for the device free step counting. Previous research of human activity sensing with commercial Wi- Fi mainly focuses on single person activity recognition. The primary challenge for the multi-person activity recognition is too difficult to derive each person's motion induced signal. All the independent running induced signals are mixed together with similar frequency and the common time-frequency analysis approaches do not work. The problem becomes even more difficult with only one pair of commodity Wi-Fi devices, which have limited number of antennas and bandwidth. In this paper, we propose Wi-Run, a multi-runner step estimation system with only one pair of commodity Wi-Fi devices. Wi-Run is composed of three innovative methods: (1) Canonical Polyadic (CP) decomposition can effectively separate running related signals. (2) The stable signal matching algorithm is applied to find the decomposed signal pairs for each runner. (3) The peak detection method is adopted to estimate steps for each runner. The multi-runner step estimation is achieved without introducing extra overhead. The experimental results illustrate the superior performance of Wi-Run, whose accuracy is about 88.25% on average.