Vision Based Finger Spelling Recognition for Ethiopian Sign Language

No Thumbnail Available



Journal Title

Journal ISSN

Volume Title


Addis Ababa University


In this paper we describe a method for automated recognition of Ethiopian Sign Language (ESL) finger spelling from a video. The method automatically selects images from a given movie. To understand the meaning of the selected images, it applied image preprocessing techniques, global thresholding, grouping neighborhood and calculating center of mass on the selected images. After applying these techniques, the method uses finite state automata to recognize the ESL finger spellings. The method recognizes the seven vowels of ESL. The method is experimented using the 238 ESL finger spelling and achieves 90.76% recognition performance, through which each of the seven vowels have 34 representation from each of the 34 consonants. As a result, the method is appropriate to recognize the ESL finger spellings integrating with the previous or future works on ESL consonant recognition. Index Terms—Sign Language Recognition, Finite State Automata and Video Understanding.



. Sign Language Recognition; Finite State Automata and Video Understanding