Wi-Run: Device-free step estimation system with commodity Wi-Fi

Abstract Quantifying running is critical to ensure its effectiveness. Current running quantification usually resorts to device-based sensing. However, the device-based step counting can not be widely adopted due to its inconvenience, especially in multi-runner running scenario. Recently, the quick development of Wi-Fi based non-invasive human activity sensing has showed a promising future for device free exercise quantification. However, existing Wi-Fi based human activity sensing systems ignore multi-object activity due to its challenge. To address this issue, we propose Wi-Run, a complete model based non-invasive step estimation system that leverages commercial Wi-Fi devices to intelligently estimate steps. Wi-Run composes two models. The first is the single runner CSI-step estimation model, which quantifies the relationship between CSI dynamics and single runner running in place steps. The second is the multi-runner CSI-step model that quantifies the relationship between CSI dynamics and each runner's running in place steps. The experimental results in typical indoor environment illustrate the superior performance of the step estimation, whose accuracies range from single runner's 93.18 % to five runners' 81.47 % on average.