基于SVM和纹理的笔迹鉴别方法

Writer Identification Using Support Vector Machines and Texture Feature

摘要: 针对与书写内容无关的笔迹,提出利用快速Gabor小波提取笔迹图像的整体纹理特征、用支持向量机(SVM)进行训练和识别的方法.SVM是解决两类问题的算法,而笔迹鉴别是一个多类问题,通过"一对多"的方法将多类问题转化为两类问题.在87人笔迹库上的实验结果表明,文中基于SVM和纹理的笔迹鉴别方法是有效的.

Abstract: A method is presented for text independent writer identification using multi-class SVM and texture feature of the whole script. 2D Gabor filter is applied to feature extraction, after that multi-class SVM is used to train and test the data. In the WS ICT writer script gallery collected from 87 persons, the proposed algorithm obtained competitive results.