Computer Science · Computer Vision and Pattern Recognition
Real-time Indoor Activity Monitoring for Elderly Using IMU-Based Signal Analysis
Ananda Garin Mills, Pakawat Panpaisan, Parin Jatesiktat, Nanthayod Termkoh, P.S. Yu, Supatana Hengyotmark, Teerayut Horanont
See on the mapAbstract
As the need for comprehensive care for the elderly grows, there is a rising emphasis on occupational management rather than solely on surveillance. This study aims to develop an advanced activity monitoring system that integrates positional data and IMU sensor data from devices positioned throughout the home environment. Previous systems have effectively addressed fundamental movements such as walking, sitting, and lying down; however, they fall short in detecting lateral and physical activities. This research seeks to expand the functionality of existing systems by enhancing the model's capabilities through targeted training, enabling it to accurately identify intricate activities such as sleeping, inactivity, extended walking, and anomalies (lying in the toilet). During test site evaluations, the system achieved 84.27% accuracy for high-level activity predictions, significantly outperforming low-level predictions at the same site, which achieved 76.29% accuracy. These advancements highlight the system's potential for real-world application, offering efficient and reliable detection of complex activities to promote greater independence for elderly individuals and reduce caregiving burdens.