Feature Extraction for Handwritten Chinese Character by Overlapped Dynamic Meshing and Fuzzy Membership
Tian Wu · Dianzi xuebao · 2004
Dynamic meshing integrates nonlinear shape normalization into feature extraction without introducing any new distortions.A feature extraction method using overlapped dynamic meshing is proposed for handwritten Chinese Character Recognition.a density image is defined for an input character image.Based on the density image,overlapped meshes and fuzzy membership are dynamically calculated.Fuzzy contour direction feature (FCDF) extracted with the proposed method is introduced.Experimental results show that significant improvement in recognition rate is yielded compared to conventional static or dynamic meshing methods.