摘要
Unrecovered falls in bedrooms can lead to severe or even fatal outcomes for the elderly population, especially those living alone. Unrecovered falls often result from neglected
care or delayed response. Timely intervention therefore becomes essential, making an effective patient monitoring system which is critical for minimizing harm. However, existing monitoring
solutions often fail to meet real needs in bedrooms because they compromise privacy, signal quality, or user compliance. In this context, this study explores the use of Frequency Modulated
Continuous Wave (FMCW) radar as a practical detector for dangerous incidents, offering a privacy-preserving alternative for home safety monitoring. Five participants modelled eight classes of bedroom activities, namely Falling, Falling off Bed, Get on Bed, Get off Bed, Flip in Bed, Sitting Down, Squatting Down, and Walking. A simple three-stage classification technique was used to process the radar data. The first stage detects the existence of Doppler signals, the second stage examines the subsequent signal following an activity, and the third stage uses an empirically determined peak velocity threshold to further identify the unrecovered falls that require attention. The preliminary results demonstrate that the proposed technique can be used
to effectively detect the unrecovered falls that requires attention.