Research Article
A Computing-Efficient Algorithm for Accelerometer-Based Real-Time Activity Recognition Systems
@INPROCEEDINGS{10.4108/icst.bodynets.2013.253623, author={Pejman Ghorbanzade and Ali Khaleghi and Ilangko Balasingham}, title={A Computing-Efficient Algorithm for Accelerometer-Based Real-Time Activity Recognition Systems}, proceedings={8th International Conference on Body Area Networks}, publisher={ICST}, proceedings_a={BODYNETS}, year={2013}, month={10}, keywords={activity recognition systems real-time and embedded systems pervasive computing}, doi={10.4108/icst.bodynets.2013.253623} }- Pejman Ghorbanzade
Ali Khaleghi
Ilangko Balasingham
Year: 2013
A Computing-Efficient Algorithm for Accelerometer-Based Real-Time Activity Recognition Systems
BODYNETS
ACM
DOI: 10.4108/icst.bodynets.2013.253623
Abstract
Considered as fundamental part of many pervasive applications, human Activity Recognition (AR) systems have recently attracted interest of the research community. One of the many challenges in developing reliable AR systems is accurate recognition of human daily physical activities while maintaining simplicity of the recognition algorithm, essential to meeting real-time functionality of AR systems as well as dealing with their processing ability constraint. In this paper, we propose a real-time computing-efficient AR algorithm for accelerometer-based AR systems. Evaluation of ¬the proposed algorithm is conducted in a laboratory setting using a simple learning based AR system with single wearable triaxial accelerometer attached to human thigh or wrist. Simple sequential human gestures are shown to be recognized with an average recognition accuracy of 98.8% and 96% for ambulatory movements and hand gestures, respectively.