Model-Based Location Tracking of an a priori Unknown Number of Inhabitants in Smart Homes

Abstract

International audienceIn this paper we propose an approach for online model-based location tracking of inhabitants in Smart Homes. Based on our previous solution for a fixed and predefined number of inhabitants using finite automata we generalize the approach to handle an a priori unknown number of inhabitants. The online-algorithms to detect the number of inhabitants and to track their location are given and a comprehensive evaluation method is defined. It consists of an analytical and a simulation-based procedure as well and is able to predict the tracking performance for real Smart Homes by a normalized measure. Throughout the paper, previous and new results are illustrated by a realistic case study and illustrative scenarios are given. Note to Practitioners—Tracking the location of several inhabitants in a Smart Home using only ambient, non-invasive, low-cost sensors is a challenge. In this paper, we propose specific evaluation approaches as well as an expert-in-the-loop improvement procedure to help the expert setting up a satisfying instrumentation using only such sensors. We also detail the algorithm for online Location Tracking of any inhabitants in a Smart Home and explain this algorithm with an illustrative scenario

An Event-Based Approach for Discovering Activities of Daily Living by Hidden Markov Models

“…12 includes only 10 actions: 1-"Watching TV", 2-"Going out", 3-"Doing laundry", 4-"Washing dishes", 5-"Preparing eat", 6-"Washing hands", 7-"Bathing", 8-"Dressing", 9-"Grooming", 10-"Toileting" and 11-"None" represents that no action was detected when one is declared. Analysing this matrix, we can conclude that actions "Watching TV"(1), "Going out"(2) and "Preparing eat" (5) are perfectly inferred by the proposed approach. On the other hands, actions "Doing laundry"(3) and "Washing dishes"(4) are non-dissociable.…”

“…Indeed, information given by this kind of sensors does not depend on the ability or the willingness of the inhabitant to provide this information. Nevertheless, information provided by binary sensors are semantically poor, what makes difficult to infer a model of the observed behaviour [5].…”

“…They are more or less intrusive, they provide more or less semantically rich information and are more or less expensive [11,12,14]. In order to help the user to accept the observation of his movements and to guarantee the respect of his privacy we chose to use non-wearable and non-intrusive sensors even if this makes multi-habitant issue harder to solve [5]. Furthermore, to reduce the costs (aging of population is a mass problem), we favour low-cost sensors.…”

“…Furthermore, to reduce the costs (aging of population is a mass problem), we favour low-cost sensors. Such sensors are mostly binary sensors (door barrier sensors, motion detectors) or sensors delivering a signal that can be interpreted as binary using a threshold (electricity consumption, water flow or pressure sensor for instance) [5]. The resulting problem for ADL is therefore to build semantically rich models allowing to know in real time which activity is performed by an inhabitant living in a smart home, from semantically poor sequences of observed binary events.…”

An Integrated Framework for Binary Sensor Placement and Inhabitants Location Tracking

“…[1]. The AAL research field aims at improving the comfort and the safety of aged or disabled people living in their own dwelling by including several applications such as the remote monitoring of patients having long-term diseases, the control of smart equipment to support elderly people in their daily activities, and the indoor location tracking of inhabitants [1]- [3], [12], [17], [18]. In particular, indoor location tracking is a meaningful evidence for numerous AAL activities.…”