سال انتشار: ۱۳۸۳

محل انتشار: سومین کنفرانس ماشین بینایی و پردازش تصویر

تعداد صفحات: ۷

نویسنده(ها):

Hadi Sadoghi Yazdi – Department of Electrical Engineering, Tarbiat Modares University, P.O.Box 14115-143, Tehran, Iran
Mojtaba Lotfizad – Department of Electrical Engineering, Tarbiat Modares University, P.O.Box 14115-143, Tehran, Iran
Mahmood Fathy – Faculty of Computer Engineering, Iran University of Science and Technology, Tehran, Iran
Ehsanollah Kabir – Department of Electrical Engineering, Tarbiat Modares University, P.O.Box 14115-143, Tehran, Iran

چکیده:

In this paper, a new type of the HMM model is presented with a recurrent kernel in the high dimensional space, HDS. HMM in the HDS have three properties: HDS is a linear space relative to the input space, creation of an interaction between HMM models, and it is a kind of hierarchical classification. Also, recurrent kernel in HMM model has a few advantages as, reducing of noise in noisy data and parameter smoothing in HMM. This approach was applied to synthetic motion and vehicle trajectory recognition; Experimental results show an increase of 32.1% for synthetic motion recognition and 9.7% for vehicle trajectory recognition performance in comparison to conventional HMM