This paper proposes a skeleton-based human behavior classification system using thinning algorithm and Support Vector Machine (SVM). The proposed system consists of two phases, skeletonization phase where main human body part is constructed using thinning algorithm, and classification phase where the skeleton constructed by previous phase is classified into certain human behavior pose using SVM. Experiment using 44 training and 44 testing data of real human poses shows that the system achieves 81.06% accuracy. This system can be further developed for early detection of criminal action.
|Number of pages
|Journal of Advanced Computational Intelligence and Intelligent Informatics
|Published - Jan 2010
- Support vector machine
- Thinning algorithm