Recognition of Gestures in Arabic Sign Language using Neural Networks

O. Al-Jarrah, A. Shatnawi, and A.H. Halawani (Jordan)


Pattern Recognition, Sign Language Recognition, Feedfor ward Neural Networks, Probabilistic Neural Networks


This paper presents two neural network systems for the recognition of the sign language alphabets. The first is based on the feedforward neural network architecture, while the second is based on probabilistic neural networks. The user is not required to wear any electromechanical de vice or any marker to interact with the system. The sys tem takes images of gestures and converts them to a set of features suitable to be fed to the neural network structure for recognition. The feature extraction scheme is robust against the changes in the hand size, orientation, and po sition within the image. This special flexibility gives the system a high degree of reliability. It is shown that an accu racy of 94.4% is achieved using the first system and 91.3% using the second.

Important Links:

Go Back