On-line handwriting recognition of Chinese characters via a rule-based approach

This paper presents a rule-based approach for online Chinese character recognition without writing constraints on both stroke number and order. Stroke correspondence is accomplished based on rules predefined such that combinatorial exhaustion can be avoided for character matching. Stroke correspondence rules contain the basic stroke types and geometric features of strokes. Because rules are constituted by invariant structural features, wide handwriting variations can be accommodated. Utilizing the structural characteristics in Chinese characters, we represent the reference database hierarchically to reduce its data size. It only requires O(n log n) time to accomplish stroke correspondence between a template of n strokes and an input script using the hierarchical reference database. This approach is very suitable for those systems with very limited computing resource.