Feature extractor giving distortion invariant hierarchical feature space

Jouko Lampinen*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

2 Citations (Scopus)

Abstract

A block structured neural feature extraction system is proposed whose distortion tolerance is built up gradually by successive blocks in a pipeline architecture. The system consists of only feedforward neural networks, allowing efficient parallel implementation. The feature extraction is based on distortion-tolerant Gabor transformation and minimum distortion clustering by hierarchical self-organizing feature maps (SOFM). Due to unsupervised learning strategy, there is no need for preclassified training samples or other explicit selection for training patterns during the training. A subspace classifier implementation on top of the feature extractor is demonstrated. The current experiments indicate that the feature space has sufficient resolution power for a small number of classes with rather strong distortions. The amount of supervised training required is very small, due to many unsupervised stages refining the data to be suitable for classification.

Original languageEnglish
Title of host publicationAPPLICATIONS OF ARTIFICIAL NEURAL NETWORKS II
EditorsSteven K. Rogers
PublisherSPIE - The International Society for Optical Engineering
Pages832-842
Number of pages11
EditionPTS 1 AND 2
ISBN (Print)0819405787
DOIs
Publication statusPublished - 1991
MoE publication typeA4 Article in a conference publication
EventApplications of Artificial Neural Networks - Orlando, United States
Duration: 2 Apr 19915 Apr 1991

Publication series

NamePROCEEDINGS OF THE SOCIETY OF PHOTO-OPTICAL INSTRUMENTATION ENGINEERS (SPIE)
Volume1469

Conference

ConferenceApplications of Artificial Neural Networks
Abbreviated title2
CountryUnited States
CityOrlando
Period02/04/199105/04/1991

Fingerprint Dive into the research topics of 'Feature extractor giving distortion invariant hierarchical feature space'. Together they form a unique fingerprint.

Cite this