INDOOR HUMAN ACTIVITY RECOGNITION METHOD USING CSI OF WIRELESS SIGNALS

Mohammed A. A. Al-qaness, Yousif Al-Eryani and Nashat Al-jallad

ABSTRACT: Human activity recognition has been studied for decades by leveraging vision-based and sensor-based technologies. However, the drawback of such techniques such as short-range area, and invasion of human privacy in vision-based technology, and inappropriate usage for sensor devices or inconvenient feeling of the user to carry sensor devices. All of these reasons encouraged researchers to wireless-based sensing technology or called device-free because the user needn’t carry devices nor monitoring with a camera. In this regard, we present a device-free activity recognition system by exploiting the Channel State Information (CSI) of Wi-Fi signals which recognizes three dynamic activities. We build our prototype with a high efficient feature extraction algorithm and an agile and accurate classification algorithm. To examine the feasibility and performance of the proposed system, hundreds of experiments have been implemented in LOS and NLOS scenarios in a dynamic environment with different volunteer users. The experiment results show that the proposed method has gained a high-accuracy rate in both LOS and NLOS.

Keywords: Human activity recognition, CSI, Wi-Fi, Device-free