Performance enhancement techniques for human action recognition in videos

Abstract

Human Action Recognition (HAR) is one of the most active research newlineareas in computer vision. It plays a vital role in wide application domains such as newlinevideo surveillance, content based video retrieval, human machine interaction, gait identification, gesture recognition, video indexing and understanding, etc. In this modern world, the safety of people is ensured and monitored by video surveillance. The advances in the field of surveillance systems have brought the technologies into reality and these technologies are applied to many real world application areas such as smart cities, parking lots, shopping malls, ATM centres, etc. The surveillance systems are manually monitored by a visual analyst and the visual analyst reports to higher authorities whenever an unusual event occurs. This sort of video surveillance analysis is man power intensive and may lead to errors. The main purpose of HAR is to automatically recognize the actions performed by a person in the videos to overcome the limitation of manual monitoring. This research work focuses on recognition of single person actions in newlinea video. It aims at enhancing the performance of HAR in videos by applying newlinerepresentative framelets selection and effective feature extraction techniques. newlineIn this research work, five approaches have been proposed for improving the newlinerecognition performance of HAR in videos. newline newline