Human Activity Recognition Using Deep Learning Approach-Application for Patient and Elderly Person Supervision

Abstract

Given the rising number of elderly individuals and growing demands on healthcare systems, Human Activity Recognition (HAR) has emerged as a vital tool for improving the care and well-being of elderly individuals and patients. By accurately identifying and classifying activities, HAR enables caregivers and medical professionals to closely monitor movements, detect anomalies, and provide timely assistance. In this study, we propose a Multi-Branch Convolutional LSTM Network (MBCLNet), a state-of-the-art deep learning model for HAR. Through a systematic approach, we develop and evaluate it using two benchmark datasets WISDM and PAMAP2. Our model achieves impressive accuracy rates of 99.98% and 99.95% on the respective datasets. By integrating MBCLNet into an API we enable real-time processing through a user interface, we enhance monitoring and supervision in healthcare environments, leading to improved control, care, and well-being for patients and elderly individuals. This research contributes to advancing healthcare outcomes and patient safety, with the potential for widespread implications in healthcare settings