Smartphone Sensor Fusion based Activity Recognition System for Elderly Healthcare

Abstract

With recent advancements in the tele-monitoring and ambient assisted living technology, human activity recognition (HAR) has proven enormously important in elderly healthcare. With the rapid increase in the use of smartphones embedded with a wide variety of latest locomotion sensors in our daily life, a new role for smartphones as the performance evaluator for physical activity recognition has emerged. HAR by using the fusion of smartphone sensors data is comparatively a new area for exploration. In this paper, we have evaluated different classification algorithms for recognition of eight physical activities performed by individuals using the smartphone tri-axial accelerometer, gyroscope and magnetometer sensors. Our analyses of collected data indicate that sensor combination improves the overall performance of the classifiers to the maximum compared to their individual performances especially for walking upstairs and downstairs activities. Moreover, we propose the use of sensor fusion for activity monitoring and diagnostic suitable for heart failure patients.