User-Dependent Sign Language Recognition Using Motion Detection

Sign language is the primary means of communication used by deaf people. Statistics show that around 360 million people around the world suffer from hearing loss. For those people, communication with hearing people is a tiring day to day process which may have an adverse effect on their lives. Sign Language Recognition (SLR) is relatively new area, it is not as mature as speech recognition for example. Moreover, Arabic Sign Language Recognition (ArSLR) did not receive much of attention until recent years. This paper presents a continuous sensor-based ArSLR system based on Hidden Markov Models (HMM) and a modified version of k-nearest neighbor (KNN). The proposed system is tested on two datasets. The first was collected using DG5-VHand data gloves and the second was collected using Polhemus G4 tracker. Each dataset was collected by a different signer. Both datasets consist of 40 Arabic sentences with 80-word perplexity. It is intended to make the collected datasets available for the research community. The proposed system provides an excellent performance of 97% sentence recognition rate.