Exploiting the wireless RF fading for human activity recognition

This paper presents a new approach to include fading effects due to human in a wireless link and exploit it for activity recognition. It is proposed as a low-cost, low complexity solution for detecting human activity by considering fading characteristics of received signal strength (RSS) at 2.4 GHz frequency on commercial IEEE 802.11 devices. The RSS mean and the fluctuation analysis (FA) are two features extracted and used for developing a recognition procedure. To evaluate accuracy performance, RSS measurements from four human activity scenarios are performed at three different environments. Collected RSS data are used for the supervise-based activity recognition scheme proposed in this paper. From the experiment, this approach can achieve on average of 90% accuracy or higher in classifying four selected human activities at separation distance of 5 m between a transmitter and a receiver. Recognition accuracy values at other environments/settings are slightly dropped due to the growth of possible unblocked multipaths, but the accuracy performance is still attained within acceptable level. The histograms or RSS distribution obtained from different human activity scenarios are also considered and analysed. Finally, the proposed approach provides decent accuracy outcomes and it demonstrates pertinence to be adopted in the future intelligent pervasive system.