Neural network Feature Extraction for the Tasks of Visual Recognition

AbstractIn this Paper, a neural network image recognition system is used. TheNeocognitron[8] in that system is used as feature extractor, then the feature areclassified by using a multilayered feedforward network to generate recognitioncodes. Many neural learning algorithms are used to extract the feature, thencomparison among them is presented. Finally a comparison between most activealgorithms among them with respect to the whole performance of the of thedesigned system is presented. The biases used in MBCL (Modified BiasCompetitive Learning) played an important role to improve the performance ofcompetitive learning algorithms. Using SOFM (Self Organizing Feature Map) toextract features gave better recognition rate than MBCL and other algorithms.Keywords: Feature extraction, Competitive learning, Image recognition,Neocognitron