In this paper, we use the depth information to effectively locate the 3D position of hands in sign language recognition system. But the information will be changed by different signers and we can’t do recognition well. Here, we use the incremental changes of the three- dimensional coordinates on a unit time as feature set to fix the above problem. And we use hidden Markov models(HMMs) as time-varying classifier to recognize the moving change of sign language on time domain. We also include HMMs with scaling factor to solve the underflow effect of HMMs. Experiments verify that the proposed method is superior then traditional one.
第二十六屆電腦視覺、圖學暨影像處理研討會暨國科會智慧計算學門成果發表及新進人員研習會=The 26th Conferenve on Computer Vision, Graphics, and Image Processing, 5p.