Hands shape recogniton using moment invariant for the Korean sign language recognition

This paper reports a recognition system of Korean sign language. The sign language is a visual language usually used by the deaf and dumb. The user expresses the meaning using the hands' motion, position and shape. This paper reports recognizing chiromancy which expresses the meaning by the hands' shape. Using the visual information, the system is able to recognize the chiromancy. The system uses spatial feature of an object in order to recognize hands' shape because the chiromancy communicates the meaning by a spatial arrangement of the shape. For analyzing the spatial feature, the moment invariants are adopted. The moment invariance is the method of the analyzing the physical property. The moment invariance is able to recognize the meaning of transformed the translation, the scale change and the rotation of hand's shape. The alphabets in the chiromancy of the Korean sign language consist of 14 consonants and 16 vowels. The chiromancy is able to distinguish the meaning according to the fingers' shape and posture. The problem often arises even when the hand has the identical shape, because the meaning changes according to the posture of hand. Therefore the principal axis is used in order to distinguish the posture to understand the meaning. Moment invariance contains abundance of information about the hands' shape. The lower order moment is not sensitive to the shape changes. Therefore, we divide the region into quadrants by the principal axes. Normalized moments of each quadrant region are used to determine the hands' shape. In the experiment, hand shapes used in Korean sign language are determined 97% in exactness.