Presence detection from smart home motion sensor datasets: a model

Abstract

Effective physical presence detection in Smart Homes (SH) can be very useful for supporting ageing in place and Ambient Assisted Living (AAL) solutions. It can contribute to ensuring peace of mind and sense of comfort for patients, their families and their carers as well as enhancing their overall physical security. Motion sensor data acquired from SH contain rich, contextual information that can be used to infer single- and multi-user presences. When combined and analysed with other information such as time of day, (expected) total number of occupants and location of any additional presence, among others, decisions can be made about whether the presences detected are legitimate or intrusive. This paper develops a presence detection model based on the analysis of a real-life SH Motion Sensor dataset. The paper also investigates the use of the relationships that exist between devices (nodes) and locations (zones) in SH. The premise is that rules can be used to constrain these relationships and, in combination, the relationships and rules can be used to describe a healthy smart home such that any deviation from the defined healthy state will result in an anomaly being flagged by a SH monitoring system.