Detecting Human Walking Direction Using Wi-Fi Signals

Using wireless signals, it is possible to map some patterns in the received signal to human activities in the vicinity of the wireless receiver. One of the potential applications of this is to detect the direction of human movement in a user device-free manner. To achieve it, we propose to use the channel state information (CSI) of received Wi-Fi signals and process the raw CSI using calibration, Hampel filter, and discrete wavelet transform to minimize the noise and retrieve the useful features of phase and amplitude components of the CSI. Then machine learning algorithms are applied to processed CSI to classify the direction of human walk. Our experimental study with off-the-shelf commodity Wi-Fi hardware and diverse users showed that the proposed system can produce 92.9%, 95.1%, and 89% accuracy for data from two different environments, combined data, respectively.