Human Activity Recognition Using BPNN Classifier on HOG Features

With an increase in number of surveillance cameras being used everywhere, there is a need of automatic systems for analysis of human activities. Human Activity Recognition is an active research area in computer vision. Research in computer vision has always focused on developing the systems capable of operating automatically. Human activity recognition has many applications such as gesture recognition, robot learning, security systems etc. Human activity recognition is achieved by human detection, feature extraction and classification of activities. In this paper human activity recognition is performed by human detection using background subtraction, feature extraction using HOG and activity classification using BPNN.