HMM-based sign recognition in consideration of motion diversity

In this paper, details are furnished with the method of sign language recognition based on hidden Markov model (HMM). It is aimed that the richer transitional topology of model would be better at accounting for motion variation modeling. In this paper, we describe a method of constructing various types of transitional topology of HMM by sharing common segments over the several sequences of sign. Thus it is shown that recognition performance for 100 isolated signs obtained based on the said method is high enough to overcome a method for which linear left-to-right HMM is used.