Isolated Word-Level Ethiopian Sign Language Recognition

No Thumbnail Available



Journal Title

Journal ISSN

Volume Title


Addis Ababa University


Hand gesture enables deaf people to communicate with each other and/or with hearing people on their daily lives rather than speaking. With regard to this there is still big communication gap between the deaf and hearing community because most of them don’t know the language. Therefore, in order to narrow this gap teaching sign language to the hearing community is one solution, however, coming up with a computerized system that can translate sign language to text or sound and vice versa is a better solution. In this regard, a lot has been done for most of the sign languages all over the world. Even though, little attention was given to the Ethiopian Sign Language (EthSL), some attempts were made to come up with a system. Attempts to develop Ethiopian Manual Alphabet (EMA) recognition system from a static image were made. As an extension to this a recognition system that can recognize continuous gestures from sequence of video frames and that also determine hand movement trajectory was proposed. However, EthSL comprises not only EMA but also gestures that represent a whole word, so a recognition system that works at that level is required. In this thesis, a system that extracts hand gestures and motion trajectory for EthSL word is proposed. The system has three modules namely the data pre-processing, feature extraction and sign classification or recognition. The preprocessing starts off with the identification of key frames, followed by skin color detection to segment hand gestures. The feature extraction module is responsible for creation of manual features and determination of hand trajectories that combines them to create a feature vector. In which the classification module trains the system as well as build a model that can be used as a reference for the recognition of a sign. The proposed system is signer dependent and experimentations are conducted using real EthSL videos. The system achieves an overall recognition rate of 40%. Keywords: Ethiopian Sign Language (EthSL); Hand Gesture; Hand Tracking; Feature Extraction; Classification.



Ethiopian Sign Language (EthSL);, Hand Gesture, Hand Tracking, Feature Extraction, Classification