The study of feature vector in HMM-based Wii application for Thai sword dance

In beginning to classify gestures, a Wii remote is used as the primary tool for collecting raw data. Next, the Hidden Markov Model method is used to classify the gestures. The performance of this classification method is reliant on the feature vector used. In this paper, we will propose an appropriate feature vector for classifying gestures used in Thai sword dancing. The feature vectors are evaluated for their accuracy of classification based on their receptivity to acceleration, velocity, and displacement.