Human Action Recognition using Prominent Camera

Abstract

Human action recognition has undoubtedly been under research for a long time. The reason being its vast applications such as visual surveillance, security, video retrieval, human interaction with machine/robot in the entertainment sector, content-based video compression, and many more. Multiple cameras are used to overcome human action recognition challenges such as occlusion and variation in viewpoint. The use of multiple cameras overloads the system with a large amount of data, thus a good recognition rate is achieved with cost (in terms of both computation and data) as the overhead. In this research, we propose a methodology to improve the action recognition rate by using a single camera from multiple camera environments. We applied a modified bag-of-visual-words based action recognition method with the Radial Basis Function-Support Vector Machine (RBF-SVM) as a classifier. Our experiment on a standard and publicly available dataset with multiple cameras shows an improved recognition rate compared to other state-of-the-art methods.