A Human Activity Recognition System using HMMs with GDA on Enhanced Independent Component Features

Human Activity Recognition (HAR) from time-sequential video images is an active research area in various applications such as video surveillance and smart homes nowadays. This paper presents a novel approach of automatic HAR based on Generalized Discriminant Analysis (GDA) on Enhanced Independent Component (EIC) features from binary silhouette information to be used with Hidden Markov Model (HMM) for training and recognition. The recognition performance using GDA on EIC features has been compared to other conventional approaches including Principle Component (PC), EIC and Linear Discriminant Analysis (LDA) on PC features where the preliminary results show the superiority of the proposed approach.