Quality-driven sensor selection in target tracking

Abstract

Sensor networks and systems are important technologies, finding their application in several domains. A major problem with sensors is that their data quality degrades over time and varies between locations and situations, due to wear and tear, interference, occlusion via random objects, power dissipation etc.We present a methodology whereby the quality of individual sensors is characterized and the resultant quality of their combined or fused quality is obtained. Our methodology is being evaluated in a health-care application wherein sensors are deployed to track elderly people with dementia in their homes. The quality of individual sensors is used as criteria in sensor selection. The proposed solution is applied successfully in monitoring indoor movement of people, and in capturing and classifying trajectories.