Poster: Continuous human activity recognition based on wifi imaging

Automatic segmentation and action recognition have been a long-standing problem in sensorless sensing. In this paper, we propose WICAR, a Wi-Fi Imaing based Continuous Activity Recognition system to solve these problems in a different way. The key idea is that: different body parts reflect transmitted signals, the receiver receives the combination of them. We separate the received signals and get the signal intensity in each direction to draft the heat map, which shows the shape of the object. To make the features easier to extract, we detects key points of human bones through the heat map. The imaging sequence contains multiple pictures recording a continuous action at different time, and we can easily separate and recognize the action based on SVM. We implement WICAR using commodity Wi-Fi devices to evaluate its performance under different environments. Experiments show that WICAR achieves an average recognition accuracy of 90%.