Towards User Activity Recognition Through Energy Usage Analysis And Complex Event Processing

Abstract

One of the key challenges related to the field of Ambient Assisted Living (AAL) is the recognition of the user's activities of daily living. Most existing approaches rely on distributed sensors, such as cameras, RFID and motion sensors. These approaches suffer from high intrusiveness for the resident, coupled with an important amount of hardware that requires maintenance. In this paper, we explore a new, low-cost and efficient solution for fine-grained activity recognition using energy consumption as input. Existing works exploiting energy sensors see the problem from an energy saving and costs reducing point of view; the originality of our work is to characterize a user's actions and activities by decomposing the total power load into a sum of loads for individual appliances. This is done using only the data from a single energy sensor located at the main electrical panel. These contributions have been implemented and tested in real live smart home prototype, using a Complex Event Processing (CEP) engine.