基于FSVM脱机手写体汉字分类识别研究  

Classified identification of off-line handwritten Chinese characters recognition based on FSVM

作　　者：朱程辉[1] 甘恒[1] 王建平[1] 

机构地区：[1]合肥工业大学电气与自动化工程学院,合肥230009

出　　处：《计算机工程与应用》2014年第23期189-193,共5页Computer Engineering and Applications

基　　金：国家实验教学示范中心项目(No.411101)

摘　　要：针对脱机手写体汉字特点,给出一种采用模糊支持向量机粗分类的方法。根据小波分解像素密度特征,利用模糊支持向量机对汉字进行粗分类。细分类识别提取外围特征,同时融合小波多网格特征,采用一对多算法进行细识别。仿真实验表明,该方法有较高识别率。Considering the features of off-line handwritten Chinese characters, this paper presents a course classification method based on FSVM（Fuzzy Support Vector Machine）. According to pixel density characteristics of wavelet decomposition, writer makes coarse classification on Chinese characters by using FSVM. On extracting peripheral features through fine classification and recognition, together with wavelet multi-grid characteristics, this paper relatively succeeds to do fine recognition by one-against-all method. The emulation test shows that the new method has a high recognition rate.

关 键 词：脱机手写体汉字 模糊支持向量机 像素密度 小波 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]