Abstract
In this paper, we study the discriminative models like CRFs, HCRFs and LDCRFs to recognize alphabet characters (A-Z) and numbers (0-9) in real-time from stereo color image sequences. To handle isolated gesture, CRFs, HCRFs and LDCRFs with different number of window size are applied on 3D combined features of location, orientation and velocity. The gesture recognition rate is improved initially as the window size increase, but degrades as window size increase further. In contrast to generative approaches such as HMMs, experimental results show that the LDCRFs are the best in terms of results than CRFs, HCRFs and HMMs at window size equal 4. Additionally, our results show that; an overall recognition rates are 91.52%, 95.28% and 98.05% for CRFs, HCRFs, and LDCRFs respectively.