Video-based continuous sign language recognition using statistical methods

This paper is concerned with a development of a video-based recognition system of continuous sign language. The system aims for an automatic signer dependent recognition of sign language sentences, based on a lexicon of 97 signs of German Sign Language (GSL). The recognition system is based on Hidden Markov Models with one model for each sign. A single video camera is utilised for data acquisition Beamsearch is employed for the recognition task. For a better result a language model is implemented, which is able to handle a-priori knowledge of the training corpus. Different results are given for a vocabular of 52 respectively 97 signs with different employed language models (Unigram and Bigram). The system achieves an accuracy of 91.8% based on a lexicon of 97 signs without a language model and 93.2% with employed Bigrams.