Probabilistic based recursive model for face recognition

Siu Yeung Cho, Jia Jun Wong

Research output: Journal PublicationConference articlepeer-review

3 Citations (Scopus)


We present a facial recognition system based on a probabilistic approach to adaptive processing of Human Face Tree Structures. Human Face Tree Structures are made up of holistic and localized Gabor Features. We propose extending the recursive neural network model by Frasconi et. al. [1] in which its learning algorithm was carried out by the conventional supervised back propagation learning through the tree structures, by making use of probabilistic estimates to acquire discrimination and obtain smooth discriminant boundaries at the structural pattern recognition. Our proposed learning framework of this probabilistic structured model is hybrid learning in locally unsupervised for parameters in mixture models and in globally supervised for weights in feed-forward models. The capabilities of the model in a facial recognition system are evaluated. The experimental results demonstrate that the proposed model significantly improved the recognition rate in terms of generalization.

Original languageEnglish
Pages (from-to)1245-1254
Number of pages10
JournalLecture Notes in Computer Science
Issue numberPART II
Publication statusPublished - 2005
Externally publishedYes
EventSecond International Confernce on Fuzzy Systems and Knowledge Discovery, FSKD 2005 - Changsha, China
Duration: 27 Aug 200529 Aug 2005

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science


Dive into the research topics of 'Probabilistic based recursive model for face recognition'. Together they form a unique fingerprint.

Cite this