A Hybrid Model for Recognizing Human Activity Using Extreme Learning

Human activity recognition (HAR) refers to the cognitive process of identifying and labeling various actions individuals perform. HAR has garnered significant attention in recent years due to its multifaceted applications in healthcare, security, human-computer interaction domains, etc. Despite several works for recognizing HAR, this still needs to be improved due to its real-time implications, complex environment, and privacy and ethical issues. This study aims to develop a model for HAR that can accurately differentiate and categorize different human actions from sequences of video images. For this purpose, a Local Binary Pattern (LBP)-based Extreme Learning Machine (ELM) is utilized. Experimental results show the quality of the proposed model over other comparing models in terms of accuracy, precision, recall, and F1-score.