Invariant handwritten Chinese character recognition using weighted ring-data matrix

A location-, scale-, and orientation-invariant handwritten Chinese character recognition system is proposed. Five invariant features are employed in this study; the main feature is just invariant to rotation, thus a scale- and translation-invariant normalization process is needed to achieve all desired invariance. Four other features are derived from three primitives: 1-fork point, corner point, and multi-fork point. To reduce matching time, preclassification is employed. A fuzzy membership function is defined according to the weighted mean ring-data matrix, number of strokes, and number of connected components to match characters. A data set was constructed from 200 handwritten Chinese characters and comprising ten different samples of each character in arbitrary orientations. Experiments were conducted with the data set to evaluate the performance of the proposed preclassification and matching methods. The average recognition rate is about 90%; we conclude that the proposed system offers a simple solution to the complex problem of invariantly recognizing handwritten Chinese characters