WiStep

Abstract

Inspired by the emerging WiFi-based applications, in this paper, we leverage ubiquitous WiFi signals and propose a device-free step counting system, called WiStep. Based on the multipath propagation model, when a person is walking, her torso and limbs move at different speeds, which modulates wireless signals to the propagation paths with different lengths and thus introduces different frequency components into the received Channel State Information (CSI). To count walking steps, we first utilize time-frequency analysis techniques to segment and recognize the walking movement, and then dynamically select the sensitive subcarriers with largest amplitude variances from multiple CSI streams. Wavelet decomposition is applied to extract the detail coefficients corresponding to the frequencies induced by feet or legs, and compress the data so as to improve computing speed. Short-time energy of the coefficients is then calculated as the metric for step counting. Finally, we combine the results derived from the selected subcarriers to produce a reliable step count estimation. In contrast to counting steps based on the torso frequency analysis, WiStep can count the steps of in-place walking even when the person's torso speed is null. We implement WiStep on commodity WiFi devices in two different indoor scenarios, and various influence factors are taken into consideration when evaluating the performance of WiStep. The experimental results demonstrate that WiStep can realize overall step counting accuracies of 90.2% and 87.59% respectively in these two scenarios, and it is resilient to the change of scenarios.