基于混合核WLS-SVR的古汉字识别(英文)  

Recognition of ancient Chinese characters based on hybrid kernel WLS-SVR

作　　者：胡根生[1] 孙莹莹[1] 徐玲英[2] 梁栋[1] 孙小棋 

机构地区：[1]安徽大学电子信息工程学院,安徽合肥230601 [2]安徽大学学报编辑部,安徽合肥230039

出　　处：《中国科学技术大学学报》2015年第4期321-328,共8页JUSTC

基　　金：Supported by the National Natural Science Foundation of China(61172127);Natural Science Foundation of Anhui Province(1408085MF121)

摘　　要：针对现有多种分类器对具有不确定字形的古汉字识别精度不高的问题,提出了一种基于混合核加权最小二乘支持向量回归(WLS-SVR)的古汉字识别算法.WLS-SVR的权重系数采用预测误差的指数衰减函数,混合核是由具有良好局域特性的小波核函数与具有良好全局特性的RBF核函数构成.在特征提取阶段,由于全局点密度与部件结构具有全局特征,而伪二维弹性网格与局部点密度具有局部特征,因此融合了古汉字的全局和局部特征.仿真实验表明,该算法具有较高的准确率与良好的鲁棒性.The shapes of ancient Chinese characters are often uncertain, which reduces the accuracy of recognition by many classifiers. To solve this problem, a new recognition algorithm combining adaptive weighted least squares support vector regression（WLS-SVR） with hybrid kernel function was proposed to recognize ancient Chinese characters. The weight coefficients of WLS-SVR decayed at a rate of the exponential function of prediction errors. The hybrid kernel was constructed using the wavelet kernel function with local properties and RBF kernel function with global properties. For feature extraction, global point density and component structure are fused with local features of pseudo 2D elastic mesh and local point density. Experiment results show the good robustness and high recognition accuracv of the proposed method.

关 键 词：古汉字识别 WLS-SVR 混合核 特征融合 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]