Fall Detection in Ambient-Assisted Living Environments Using FMCW Radars and Deep Learning

The global rise in the elderly population has increased the demand for effective fall detection in Ambient Assisted Living (AAL) environments. This paper introduces a novel and reliable fall detection system utilizing frequencymodulated continuous wave (FMCW) radar, designed to address privacy concerns, operate reliably in low-light conditions, and provide ease of installation. Data from two wall-mounted radars capture a variety of activities, including simulated falls, across five configurations to enhance model generalizability. Radar data processing employs the Fast Fourier Transform (FFT) and the Capon algorithm to generate Range-Azimuth and Range-Elevation maps, which serve as input features for a proposed 3D Convolutional Neural Network (3D CNN) model. This model achieves an accuracy of 94.33% and F1-score of 93.5%, combining high performance with adaptability across diverse environments and user needs. This work provides a robust solution for fall detection with significant potential for deployment in real-world elderly care settings.