Cloud-based recognition of complex activities for ambient assisted living in smart homes with non-invasive sensors

Abstract

Automatic recognition of complex activities can aid in finding correlations between the daily habits of people and their health state, and can further lead to early detection of diseases or accidents. In this paper we propose a cloud-based system for recognition of complex activities by detecting series of atomic actions with non-invasive sensors. Collected data from non-invasive, non-intrusive and privacy preserving sensors is streamed into a cloud-based system, where automated feature extraction and activity recognition is performed. The prototype of the proposed system is evaluated with an experiment. Five activities performed by a person in a room were monitored by a sensor kit and streamed to the cloud, where the built classification models could recognize the activities with accuracy of 80% to 95%, depending on the length of segmentation windows which varied from 5 to 20 seconds, respectively.