Appearance-Based Hand Sign Recognition from Intensity Image Sequences

In this paper, we present a new approach to recognizing hand signs. In this approach, motion recognition (the hand movement) is tightly coupled with spatial recognition (hand shape). The system uses multiclass, multidimensional discriminant analysis to automatically select the most discriminating linear features for gesture classification. A recursive partition tree approximator is proposed to do classification. This approach combined with our previous work on hand segmentation forms a new framework which addresses the three key aspects of hand sign interpretation: hand shape, location, and movement. The framework has been tested to recognize 28 different hand signs. The experimental results show that the system achieved a 93.2% recognition rate for test sequences that had not been used in the training phase. It is shown that our approach provide performance better than that of nearest neighbor classification in the eigensubspace.