V.N. Pashaloudi and K.G. Margaritis (Greece)
Greek Sign Language, Hidden Markov Models
The work presented in this paper aims at developing a system that could recognize both isolated and continuous Greek Sign Language (GSL) sentences. Feature vectors are extracted from images, representing GSL letters and are used in sequences as input to Hidden Markov Mod els (HMMs). The feature extracting method used, derive geometric properties of the hand morph. Tests on isolated letter recognition and recognition of sequences of letters have been performed, with recognition rates reaching 90% and 95% respectively. Considering whole words as the ele ments under recognition, experiments on isolated and con nected word recognition have been performed. These ex periments concern a 26 sign vocabulary, and gave out the promising recognition rates of 97% and 86% respectively.
Important Links:
Go Back
