Enhancing Elderly People's Quality of Life by Identifying Their Actions using an Effective Machine Learning Model

Abstract

Globally, millions of people are above 65 years of age which affects both their physical and emotional well-being. They are likely to face incidents like falls resulting in potentially severe repercussions including hospitalizations.  Human Action Recognitions (HARs) are techniques that recognize and categorize human behaviours from sensor data using machine learning (ML) algorithms. These are essential for human behavior analyses in applications including diagnosis of severe illnesses, patient rehabilitations, and healthy lifestyles. One increasingly effective use of ML methods are predictions of human behaviours, whereby computers monitor routines and intervene in crisis, which may be quite beneficial for the elderly. Moreover, there aren't many studies on HARs for the elderly making it imperative to study human gestures. This work introduces a Logistic Regression Based Identification of Human actions (LRBIHA) schema based on ML that can identify human actions with accuracies of over 90%. Solutions based on LRBIHA HARs can enable the appropriate use of supported living applications for humans including home based monitoring. The schema cam also be useful to clinicians, potential therapeutics and research.