AI based Activity Monitoring in Washrooms using Low-Resolution Radar

This paper presents a low-cost radar monitoring system for comprehensive washroom activity analysis in elder care homes. Leveraging deep learning, this system can overcome the low resolution of inexpensive off-the-shelf radar to monitor residents with high accuracy. Radar data, in the form of Range-Elevation, Range-Doppler, and Range-Azimuth, is processed through a convolutional long short-term memory (ConvLSTM) neural network architecture for activity classification and anomaly detection. Our methodology achieved an accuracy over 90% in identifying routine behaviours and fall events for 5 participants.