Streaming Processing for ADL Monitoring in Smart Home Environments

Abstract

Monitoring and detection of activities of daily living (ADL) is a frequent practice to determine the independence of elderly/disabled people in their homes instrumented with a set of sensors, which conform smart home environments. In these scenarios there are two main problems that need to be addressed: the representation of the ADL and the real time processing of the data gathered by the sensor network. Regarding the first aspect, a reliable representation and modeling of ADL to support their automatic representation must consider factors such as human location, presence of physical objects, and time. Concerning the second aspect, the huge volume of data produced by the sensor network at different velocities and with varied formats must be processed as they are generated in order to be able to provide timely responses. In this sense,  we extend a previously proposed framework aimed at assessing the level of independence of an elder person living in a smart home by integrating: (i) the HAMSTERS-XL notation to represent ADL and to instantiate independence evaluation models; and (ii) capabilities for batch and streaming processing, based on Big Data engines. We illustrate the suitability and functionality of the extended framework with a use case consisting of a virtual smart home environment and an inhabitant performing five ADL represented with HAMSTERS-XL and the AGGIR grid model. With this experience, we highlight the benefits of task models to represent ADL combined with the use of Big Data tools.

Keywords: IoT sensor networks, ADL, pervasive health systems and services, AGGIR grid, Big Data, smart homes