Multi-Modal Nonlinear Feature Reduction for the Recognition of Handwritten Numerals

Authors:
Affiliations:
Venue:
CRV '04 Proceedings of the 1st Canadian Conference on Computer and Robot Vision
Year:
2004

Citing 5
Cited 0

Introduction to statistical pattern recognition (2nd ed.)

Introduction to statistical pattern recognition (2nd ed.)
Floating search methods in feature selection

Pattern Recognition Letters
Feature selection using a proximity-index optimization model

Pattern Recognition Letters
Off-Line Recognition of Totally Unconstrained Handwritten Numerals Using Multilayer Cluster Neural Network

IEEE Transactions on Pattern Analysis and Machine Intelligence
A Branch and Bound Algorithm for Feature Subset Selection

IEEE Transactions on Computers

Quantified Score

Hi-index	0.00

Visualization

Abstract

A novel method of multi-modal nonlinearfeature reduction is proposed for therecognition of handwritten numerals. In orderto find an effective decision boundary, eachclass is divided into several clusters. Then thek-NN sorting algorithm is applied to eachcluster to get the training data along theeffective decision boundary. Optimaldiscriminant analysis is implemented by multimodalnonlinear mapping to generate abetween-class scatter matrix, which requiresless CPU time than other nonparametricapproaches. Experiments demonstrated thatour proposed method could achieve a highfeature reduction without sacrificing muchdiscriminant ability. As a result, this newmethod can reduce ANN training complexityand make the ANN classifier more reliable. Itsfeature dimensionality reduction outperformsthe PCA and mono-modal nonparametricanalysis.