Handwritten Kanji recognition with the LDA method

We describe the application of the LDA method (learning by discriminant analysis) to handwritten Kanji recognition. In the LDA method, the discriminant function obtained by applying Fisher´s linear discriminant analysis is superposed onto the original distance function such as the weighted Euclidean distance. The LDA method is applied to transformed features obtained by using canonical discriminant analysis. We also propose a new feature extraction method that reflects structural information of Kanji. When the LDA method is used, the misrecognition rate for test patterns was reduced by 36%