Development of a portable activity detector for daily activity recognition

Abstract

This paper presents a portable device for real-time daily activity identification. The proposed portable device is realized by an embedded system that integrates a triaxial accelerometer, a microprocessor, and a wireless transceiver module. An online activity recognition algorithm has been developed for human daily activity recognition based on the acceleration signal collected from the triaxial accelerometer. The proposed algorithm is composed of data collection, data preprocessing, feature extraction, feature reduction, and classifier construction. Our experimental results have successfully validated the effectiveness of the portable activity detector in terms of high accuracy.