Human Activity Detection via WiFi Signals Using Deep Neural Networks

This study proposes a WiFi-based activity detection system using deep neural networks to detect the indoor human states. This system captures useful amplitude information from the channel state information and converts the information to two-dimensional arrays. Next, the two-dimensional arrays are used as inputs to deep neural networks to distinguish the moving and stationary states of people. The powerful inference of deep neural networks simplify the feature extraction and also improve the accuracy of the classification of indoor human states. Our prototype shows that the proposed system can work with WiFi signals with even higher accuracy.