Hand-written Japanese Kanji character recognition by a structured self-growing neural network “CombNET-II”

Abstract

CombNET-II is a self-growing 4 layered neural network model which has a comb structure. The first layer constitutes a stem network which quantizes an input feature vector space into several sub-spaces and the following 2-4 layers constitutes branch network modules which classify input data in each sub-space into specified categories. The excellent performance of CombNET-II is demonstrated using hand-written Japanese Kanji characters(1,000 categories). 768 peripheral direction contributivity (PDC) features extracted from each Kanji character are used as input data for CombNET-II. CombNET-II correctly classified 96.8% of previously unseen hand-written Kanji characters.