Face recognition using hidden Markov model supervectors

This project attempts to boost the results of face recognition algorithms already established to perform face recognition by augmenting the architecture and using HMM-based supervector classification. In this thesis, the work of Tang’s 2010 dissertation is used such that the HMM based classifier takes on a UBM-MAP adaptation based approach. In addition, Tang’s work is extended to the case of pseudo 2-dimensional HMMs. Thus, a supervector classifier for pseudo 2DHMMs is developed and then applied to the task of face recognition. When the recognition algorithm is applied to the ORL database, the results show that the algorithm is able to either perform as well as other face recognition algorithms applied to this database, or actually outperform them.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Soberal, Daniel

Contributors dc:contributor

• Hasegawa-Johnson, Mark A.

• Hidden Markov Model (HMM)
• supervectors
• Gaussian mixture models
• Kullback-Leibler divergence

Statement dc:rights

• Copyright 2014 Daniel Soberal

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/72839

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/72839