Deep Learning-Based Big Data Analytics Model for Activity Monitoring of Elderly People

Abstract

AbstractHuman activity recognition (HAR) has grown in popularity over recent years, owing to the recent advancements in communication and wearable sensors. In order to handle the assemblage of big data and to analyze in real-time, deep learning is vital to extract useful information from complex systems. Deep learning-based HAR is one of the most promising assistive technology tools for supporting elderly people in their daily lives by monitoring their cognitive and physical function. In particular, fall detection is a major challenging issue since elderly people are more likely to fall. Therefore, a dedicated monitoring system is highly desirable in order to improve independent living. In this paper, a video-based fall detection system in an indoor environment is created using a convolution neural network (CNN). It also describes the design of a wearable-based fall detection system that uses an accelerometer and a gyroscope as motion sensors to detect body rotation and movement. The development of artificial neural network (ANN) for fall detection, employing wearable sensors is also investigated in this research. Experimental analysis on the ‘UR fall detection dataset’ proved the significance and robustness of the proposed approach in terms of accuracy and precision.KeywordsActivity recognitionDeep learningElderly health careBig dataOptical flow images