SEAbIRD: Sensor activity identification from streams of data

Abstract

Active Aging is a proposal that aims to improve life quality, as a person grows old. One of the main use cases of this concept is the application of products and services based on technology; this approach, known as Ambient Assisted Living (AAL). An important activity performed by AAL is the discovery of the user’s activities of daily life (ADL) employing data retrieved from sensors set on an active home. Still, there is no much research on implementing a system for ADL discovery which contemplates factors as personalized configuration, sensor failure and user privacy. We identify the main requirements that an ADL discovery system must have. Then, we propose an ADL discovery schema that supports these necessities. Finally, we explore the application of adaptable and sensor-failure tolerant ADL discovery models over recorded data from a real user. This exploration evidences that our proposed models can adapt to the above-mentioned scenarios and still have an outstanding performance on activity discovery process.