WiFi Sensing for Presence Detection Using Deep Learning

The rapid advancement of Artificial Intelligence (AI) and Wi-Fi technology have enabled the development of smart infrastructure systems. Wi-Fi sensing refers to the noninvasive process of detecting human motion in indoor environments. This is achieved by exploiting the disturbances in the Wi-Fi signal caused by human motion. This technology enables a range of applications in security, healthcare monitoring and home automation, while offering a nonintrusive and cost-efficient solution. This paper focuses on device-free human presence detection using Channel State Information (CSI). The proposed framework leverages IoT devices to design a presence detection system using Deep Learning (DL) techniques including CNN and LSTM. Experimental results show that the proposed DL models tested in different environments achieved a robust generalization in both line-of-sight (LoS) and non-line-of sight (NLOS) scenarios, especially in through-the-wall scenario. On average, the proposed CNN model trained on periodogram of the CSI amplitude outperforms the other models and achieves an accuracy of 94%.