Feature Extraction for Trajectory Representation of Sign Language Recognition

Abstract

Gestures are the dynamic movements of hands within a certain time interval, which are of practical importance in many areas, such as human computer interaction, computer vision, and computer graphics. This paper deals with the problem of representing and classifying trajectories of sign language. A proposed method of features extraction is introduced that includes time serial combined with chain code. The produced feature vector is used as input to a recurrent neural network (RNN) for recognition. The dynamic behavior of the RNN used to categorize input sequences into classes. The results show the effectiveness of the proposed method, where the input to the system is weekday’s signs and the output is the corresponding word.