부산대학교 도서관

Wi-Fi sensing technology is being used extensively for different sensing applications, mostly for human activity recognition in the recent past. The Channel State Information (CSI) identifies the frequency shifts in the wireless medium caused by movements and changes in the region of interest, which can be analyzed using the amplitude data of various frequency channels. However, the existing systems are not suitable for real-time implementation due to the three layer cloud based architecture used. With IoT devices, edge computing provides a suitable solution with negligible communication latency and reduced network traffic. The proposed work mainly focused on real-time human activity information extraction in smart home environments using CSI values of low-cost ESP32 WiFi device. The center point of this work is to implement the IoT and edge layers of the three layered architecture to extracts, processes and visualize the sensing data in real-time with low-latency. We use simple statistical features and light weight machine learning algorithms for human activity recognition in real-time instead of complex and computationally heavy algorithms which is not suitable for edge computing.