Human Activity Recognition Based on Motion Projection Profile Features in Surveillance Videos Using Support Vector Machines and Gaussian Mixture Models

Human Activity Recognition (HAR) is an active research area in computer vision and pattern recognition. The area of human activity recognition, attention consistently focuses on changes in the scene of a subject with reference to time, since motion information can sensibly depict the activity. This paper depicts a novel framework for activity recognition based on Motion Projection Profile (MPP) features of the difference image, representing various levels of a person’s interaction. The motion projection profile features consist of the measure of moving pixel of each row, column and diagonal (left and right) of the difference image and they give adequate motion information to recognize the instantaneous posture of the person. The experiments are carried out using UT-Interaction dataset (Set 1 and Set 2), considering six activities viz (handshake, hug, kick, point, punch, push) and the extracted features are modeled by Support Vector Machines (SVM) with RBF kernel and Gaussian Mixture Models (GMM) for recognizing human activities. In the experimental results, GMM exhibit effectiveness of the proposed method with an overall accuracy rate of 93.01 % and 90.81 % for Set 1 and 2 respectively, this outperforms the SVM classifier.