Gesture classification and recognition using principal component analysis and HMM

Abstract

In this paper, we describe the method that can automatically compose gesture models and recognize those gestures using 2D features extracted from gesture image sequences. In the conventional gesture recognition algorithms, previously well-known patterns are introduced by the hand or the model indexing algorithm. However, our method automatically composes the model space by clustering arbitrary input image sequences. The models are recognized as gesture using probability calculation of HMM. Our method can compose the models fast and robustly and is easy to learn on new image sequences.