Intelligent Recognition of Activities of Daily Living for Assisting Memory and/or Cognitively Impaired Elders in Smart Homes

Introduction

Improvement of life quality in the developed nations has systematically generated an increase in the life expectancy. Nevertheless, the increasing number of elderly person implies more resources for aftercare, paramedical care and natural assistance in their habitats. The situation is further complicated if elders suffer from memory disorders (Pigot et al., 2006). In this case a permanent assistance is necessary wherever they are. During the last decade, some researches (Boger et al., 2006; Mihailidis et al., 2004; Snoek et al, 2008; Tam et al., 2006) have proposed intelligent systems to assist elders with memory troubles to carry out complex daily activities. To maximise their efficiency, such systems require continuous identification of what the memory impaired person makes, recognising its intentions and analysing the tasks partially carried out; in order to help him/her (if need arise) to achieve and finalise what is already undertaken. However, the major phase of this assistance process is the recognition of what really occurs within the habitat; since the intelligent system reaction – in term of optimal planning of tasks to be made or terminated – for better assisting the occupant is based primarily on what does this system “think” that is happening.

This article describes a recognition approach of undertaken activities of daily living (ADLs) performed by memory and/or cognitively impaired elders in smart homes. The proposed technique is materialised via a recognition module inserted in a modular generic architecture which aims to offer a framework to conceive intelligent ADLs assistance systems. The remainder of the article is organised as follow. Section 1 presents the modular architecture and its components. Although the main subject of the article is the recognition module (RM), the goal of this section is to enlighten the reader about the context in which the RM is used, the stochastic model of the recognition in the RM, as well as the interaction between the various modules of the modular architecture. Section 2 is dedicated to the experimental validation where empirical results of practical tests are presented and interpreted. In section 3, we discuss the obtained results and draw some deductions about the key features that act upon the efficiency of the ADLs recognition. In the last section, by way of conclusion, we sum up this research work.