HMM based automatic Arabic sign language translator using Kinect

Abstract

In this paper, a new Arabic sign language automatic translator is presented. The translator is based on Hidden Markov Models (HMM's). The features used in recognition are 3D information detected using Microsoft Kinect Sensor depth camera. The system was trained to recognize 40 signs from standard Arabic sign language. A go-stop scheme is presented to handle sequences of signs which construct sentences in real-time. The recognition success rate based on the new methodology is above 90 percent with real time performance on a PC.