SVM-Based Human Action Recognition and Its Remarkable Motion Features Discovery Algorithm

Motivation, Problem Statement, Related Works This paper proposes a discovery algorithm of knowledge of remarkable motion features in daily life action recognition based on Support Vector Machine. The main characteristics of the proposed method are 1)basic scheme of the algorithm is based on Support Vector Learning and its generalization error, 2)remarkable motion features are discovered in response to kernel parameters optimization through generalization error minimization. Experimental results show that the proposed algorithm makes the recognition system robust and finds remarkable motion features that are intuitive for human.. Recognizing human actions has potential to contribute to intuitive communication between human and machine, such as human-computer interaction, search engine for multi-media databases, or intelligent video editing. It may be applied to design some level of humanoid actions efficiently. It is proper to divide the process of action recognition into the following two phases. The former is to get time series of 3D body motion structurally from some instruments. The latter is to symbolize these kinds of motion to action names. There are many researches which use video image as input. For example, Starner et al. constructed a sign language recognition based on HMM[1], and Wilson et.al made a gesture recognition system[2]. However, usually the main work of such researches is on how to acquire motion robustly, since the time series image processing is still a difficult problem.