Telemonitoring of Daily Activity Using Accelerometer and Gyroscope in Smart Home Environments

Abstract

Wearable sensors in the smart home environment have been actively developed as assistive systems to detect behavioral anomalies. Smart wearable devices incorprated into daily life can respond immediately to anomalies and process and dispatch important information in real-time. Artificially intelligent technology monitoring of the user's daily activities and smart home ambience is especially useful in telehealthcare. In this paper, we propose a behavioral activity recognition framework which uses inertial devices (accelerometer and gyroscope) for activity detection within the home environment via multi-fused features and a reweighted genetic algorithm. The procedure begins with the segmentation and framing of data to enable efficient processing of useful information. Features are then extracted and transformed into a matrix. Finally, biogeography-based optimization and a reweighted genetic algorithm are used for the optimization and classification of extracted features. For evaluation, we used the "leave-one-out" cross validation scheme. The results outperformed existing state-of-the-art methods, achieving higher recognition accuracy rates of 88%, 88.75%, and 93.33% compared with CMU-Multi-Modal Activity, WISDM, and IMSB datasets respectively.