Building Compact MQDF Classifier for Off-line Handwritten Chinese Characters by Subspace Distribution Sharing

Quadratic classifier with modified quadratic discriminant function (MQDF) has been successfully applied to recognition of handwritten characters to achieve very good performance. However, for large category classification problem such as Chinese character recognition, the storage of the parameters for the MQDF classifier is huge which makes it not practical to be used in hand-held devices. In this paper, we propose a method to build a compact MQDF classifier for handwritten Chinese characters. We use subspace distribution sharing to form a codebook for the subspaces of the MQDF classifier´s parameters. Further more, we employ a two-level Euclidean distance classifier to accelerate the recognition process. Fast recognition speed and compact dictionary size make the high accuracy quadratic classifier be practical for hand-held devices.