Structured classification has been adopted to recognize handwritten numerals by extracting feature primitives that characterize each image. We propose a handwritten numeral recognition system based on reduced features extraction and fuzzy membership functions, with the intention to find a minimal set of feature primitive without sacrificing the recognition rate. We first perform preprocessing of smoothing and thinning to obtain a skeleton for each image. For each skeleton, the following feature points are detected: terminal intersections, and directional. We then extract the following five feature primitives for each skeleton: loop, horizontal, vertical, C-like curve, and D-like curve. Two fuzzy S-functions are used as membership functions to estimate the likelihood of these primitives being close to the top and to the bottom of the image. A tree-like classifier based on the feature primitives and fuzzy membership is then applied to recognize the numerals. Handwritten numerals in NIST Special Database 19 are recognized with 88.72% correct rate.
二00二年國際計算機會議論文集(II)=Proceedings of the 2002 International Computer Symposium (Volume II)，頁1705-1711