Implementing Passive Wi-Fi Radar for Fall Detection for the Elderly

Falls among the elderly pose serious health risks, often leading to injuries and loss of independence. Traditional fall detection methods, such as wearable sensors and camerabased systems, suffer from limitations, including privacy concerns, user discomfort, and compliance issues. While radarbased fall detection is gaining traction, current AI-driven methods demand extensive training data and high processing power. This paper explores a simpler, more efficient approach: threshold-based classification. Its low complexity, transparency, and minimal computational requirements make it ideal for real-time fall detection and rapid notification. A dataset of fall and non-fall signals is collected from participants performing both single and dual activities. The system evaluates using threshold-based classification techniques to differentiate falls from normal movements, which are time characteristics, voltage drop, and signal angle. The findings reveal that angle-based thresholding achieves the highest performance, with $100 \%$ detection accuracy for single activities and $\mathbf{7 5. 0 \%}$ accuracy for more complex dual activities. This significantly outperforms time and voltage parameters in distinguishing falls from non-falls. These findings underscore the potential of Passive Wi-Fi Radar as an efficient, privacypreserving solution for real-time fall detection, eliminating the need for wearable devices while ensuring timely intervention.