Presentation 2003/2/14
The Training Algorithm based on Variational Approximation for Separable 2D-HMM
Yasushi OHNO, Yoshihiko NANKAKU, Keiichi TOKUDA, Tadashi KITAMURA, Zoubin GHAHRAMANI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In image recognition, a variety of elasticity, rotation and shift of image affects recognition performance seriously. To solve this problem, several methods based on 2D-HMM have been proposed. However, since the structure of 2D-HMM is more complex than that of 1D-GMM, high computational cost is required to calculate the likelihood and estimate the model parameters. In this paper, we define separable 2D-HMM which consists of two state sequences : vertical and horizontal sequences, and propose a training algorithm for separable 2D-HMM based on variational approximation. The proposed method can reduce the computational complexity as compared to N-best approximation and the lower bound of log-likelihood is guaranteed to increase at each iteration of the training algorithm
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Hidden Markov Models / 2D-HMM / Variational Approximation / EM Algorithm / Image Recognition
Paper # PRMU2002-211
Date of Issue

Conference Information
Committee PRMU
Conference Date 2003/2/14(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) The Training Algorithm based on Variational Approximation for Separable 2D-HMM
Sub Title (in English)
Keyword(1) Hidden Markov Models
Keyword(2) 2D-HMM
Keyword(3) Variational Approximation
Keyword(4) EM Algorithm
Keyword(5) Image Recognition
1st Author's Name Yasushi OHNO
1st Author's Affiliation Dept. of Intelligence and Computer Science, Nagoya Institute of Technology()
2nd Author's Name Yoshihiko NANKAKU
2nd Author's Affiliation Dept. of Intelligence and Computer Science, Nagoya Institute of Technology
3rd Author's Name Keiichi TOKUDA
3rd Author's Affiliation Dept. of Intelligence and Computer Science, Nagoya Institute of Technology
4th Author's Name Tadashi KITAMURA
4th Author's Affiliation Dept. of Intelligence and Computer Science, Nagoya Institute of Technology
5th Author's Name Zoubin GHAHRAMANI
5th Author's Affiliation Dept. of Intelligence and Computer Science, Nagoya Institute of Technology
Date 2003/2/14
Paper # PRMU2002-211
Volume (vol) vol.102
Number (no) 652
Page pp.pp.-
#Pages 6
Date of Issue