Fall Detection of Elderly in Ambient Assisted Smart Living Using CNN Based Ensemble Approach

Abstract

In recent years, there has been a significant increase in ambient assisted living and smart environment homes that utilize a range of technologies to enhance the quality of life for elderly people. Fall detection is an essential service that smart home healthcare can provide as falls can pose a significant threat to the independence and health of individuals over 65 years old. This article introduces the ISBFD (Inertial Sensor Based Fall Detection) concept, which aims to identify elderly persons who fall and alert family members or carers right away. The proposed model employs data from the accelerometer sensor of a smartphone in real-time. This data is then processed by a fall detection system that can run directly on the device. In this study, an initial-level deep learning model for fall detection is deployed along with subsequent models using ensemble learners, and it is trained on the publicly accessible MobiAct dataset. A comparative analysis is drawn between initial (Convolution Neural Networks) and final predictors (Ensemble Learners). The health and well-being of elderly people can be considerably improved by the ISBFD model, which makes it possible to detect falls and promptly warn carers with accuracy upto 93% approximately.