Recognition of sign language motion images

摘要

This paper describes a method of classifying single view deaf-and-mute sign language motion images. We suppose the sign language word is composed of a time sequence of units called cheremes. The chereme is described by handshape, movement, and location of the hand, which can be said to express the 3-D features of the sign language. First, a dictionary for recognizing the sign language is made based on the cheremes. Then, the macro 2-D features of the location of a hand and its movement are extracted from the red component of the input color image sequence. Further, the micro 2-D features of the shape of the hand are also extracted if necessary. The 3-D feature descriptions of the dictionary are converted into 2-D image features, and the input sign language image is classified according to the extracted features of the 2-D image.