Activity Recognition Using Machine Learning in Smart Spaces

A huge societal need is felt to address the problem of the growing elderly population affected with Alzheimer's disease. The proficiency with which such affected population do their Activities of Daily Living (ADL) a.k.a. Activity Recognition (AR), is of paramount importance to the medical fraternity for proper diagnosis, suggestions for treatment and necessary follow-up. AR is based on ambient sensors, which is an active research field with wide cross-domain applications. Over decades there has been a lot of research efforts made in Text Categorization Paradigm (TCP), using machine learning techniques. In this work, a novel model, called Activity Recognition - Text Categorization Paradigm (AR-TCP), is proposed to view the task of human Activity Recognition, as a special case of TCP. This work uses techniques from Bayesian theory, non-parametric classification and adapts set theoretic advances from fuzzy sets, rough sets, granular rough-fuzzy theory; with the sole aim to show the ease of extending the proposed AR-TCP model to include latest machine learning advances from the text categorization paradigm, to further the progress in the dynamic AR domain.