Human Activity Recognition Using Hierarchical Hybrid Multi-CNN ELM Classifier

Abstract

The issue of human activity detection is a complex task. A wide range of approaches to detecting human activity have been recorded and documented within the scientific community. Human activity detection is linked to a multitude of technology-driven systems that are fundamental to daily existence, encompassing human-computer interaction, surveillance systems, healthcare monitoring, video surveillance, robotics, surveillance systems, and content-based data retrieval. Traditional methods for detecting activities often rely on manually constructed attributes and rule-based systems, which may possess constraints regarding applicability and scalability. The advent of machine learning and deep learning techniques has precipitated a significant shift towards data-centric technologies capable of independently extracting unique attributes from raw sensor data. This paper provides the use of Hierarchical Hybrid Multi-CNN ELM (HHMConvELM) hybrid classifier for HAR, aiming to develop a model that outperforms single classification models on HAR datasets