WiCare: Accidental Fall Detection for Elderly Care Using Passive Wi-Fi Sensing

Monitoring people’s activities serves various purposes, with accidental fall detection being a significant one. Existing technologies rely on wearable sensors, which are only sometimes practical, and privacy concerns surround camera-based solutions. This study introduces a non-invasive approach using passive Wi-Fi sensing to identify accidental falls, offering a way to detect people and devices in an area without additional sensors. The system employs deep neural networks to detect falls by analyzing Wi-Fi signal fluctuations caused by human movement. A substantial dataset of Wi-Fi signals collected during five activities (fall, pickup, sit down, stand up, and walk) in a smart home environment trains the deep learning model. Amplitude and phase, being the fundamental properties of a signal, are used as components in feature selection for our data set. We employed Long Short-Term Memory(LSTM) and random forest algorithms to classify five activities. Our approach yielded a remarkable average accuracy of 98%, demonstrating its ability to identify and differentiate these activities based on WiFi signals precisely. This highlights the promising potential of WiFi sensing technology for practical applications in activity recognition and monitoring.