Learning Quadratic Discriminant Function for Handwritten Character Classification

For handwriting recognition integrating segmentation and cluss$icution, the underlying classifier is desired to give both high uccuruc~ and resistance to outlil3-S. In a previous evaluation study, the modified quadratic discriminunt function (MQDF) proposed by Kimuru et al. was shown to be superior in out&r rejection but inferior in cluss$icution uccuruc~ us compared to neural cluss$ers. This paper proposes a learning quadratic discriminunt function (LQDF) to combine the advantages of MQDF and neural cluss$ers. The LQDF achieves high uccurucy and outlier resistance via discriminative learning and adherence to Gaussian density assumption. The ejjicucy of LQDF was justified in experiments of handwritten digit recognition.