Abstract
Sign language is based principally on hands gestures. To have a robust recognition system, each hand modality must be correctly presented using both motion and shape descriptors. This paper proposes an enhanced isolated word sign language recognition system based on hands trajectories analysis able to solve the most challenges such as signer's interchangeability and speed variation. In this context, Histogram of Oriented Gradients (HOG) and Support Vector Machine (SVM) are introduced. The performance of our proposed system is tested on public databases (RWTH-Boston-50 and RWTH-Boston-104) with signer-independent condition and outperformed the recent existing works.