Presentation 2010-06-25
Face recognition based on extended separable lattice HMMs
Keisuke KUMAKI, Yoshihiko NANKAKU, Keiichi TOKUDA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In image recognition systems, it needs to deal with geometrical variations of an object, e.g. location, size, rotation and etc. Separable lattice hidden Markov models (SL-HMMs) have been proposed as probabilistic models including the normalization process for such variations. SL-HMMs have horizontal and vertical state sequences and makes it possible to model invariances to the size and location of an object. However, SL-HMMs cannot deal with rotation and distortion. This paper proposes extended separable lattice HMMs which have state sequences corresponding to each row and column of an input image. By complicated state alignments due to increasing the state alignments, the proposed models can deal with rotation and distortion. We also derive a training algorithm based on the variational EM algorithm and evaluate the effectiveness of the proposed model in face recognition experiments.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) hidden Markov model / face recognition / separable lattice HMM / EM algorithm / variational approximation
Paper # PRMU2010-46,HIP2010-35
Date of Issue

Conference Information
Committee PRMU
Conference Date 2010/6/17(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) Face recognition based on extended separable lattice HMMs
Sub Title (in English)
Keyword(1) hidden Markov model
Keyword(2) face recognition
Keyword(3) separable lattice HMM
Keyword(4) EM algorithm
Keyword(5) variational approximation
1st Author's Name Keisuke KUMAKI
1st Author's Affiliation Dept. of Computer Science and Engineering, Nagoya Institute of Technology()
2nd Author's Name Yoshihiko NANKAKU
2nd Author's Affiliation Dept. of Computer Science and Engineering, Nagoya Institute of Technology
3rd Author's Name Keiichi TOKUDA
3rd Author's Affiliation Dept. of Computer Science and Engineering, Nagoya Institute of Technology
Date 2010-06-25
Paper # PRMU2010-46,HIP2010-35
Volume (vol) vol.110
Number (no) 97
Page pp.pp.-
#Pages 6
Date of Issue