一种三维空间手写数字的融合识别方法  

Fusion Recognition Method on 3D Space Handwriting Digit

作　　者：曹磊[1] 

机构地区：[1]淮北师范大学计算机科学与技术学院,安徽淮北235000

出　　处：《淮北师范大学学报（自然科学版）》2013年第4期56-60,共5页Journal of Huaibei Normal University：Natural Sciences

基　　金：安徽省高等学校省级优秀青年人才基金项目(2012SQRL078);安徽省教育厅自然科学基金项目(KJ2012Z354)

摘　　要：在三维空间手写数字识别研究中,特征的选取和融合直接决定最终的识别率.空间手写数字特征一般包括时域特征和频域特征两大类,时域特征又包括加速度相关特征、旋转特征及短时能量特征;频域特征又包括DCT变换特征及FFT变换特征.针对这一特点,文章提出一种融合多种特征的识别方法.首先,对三维空间手写识别的各种子特征进行提取;然后选取最能反映字符特点的两种子特征进行融合并对融合后的特征向量进行降维处理;最后进入分类器进行识别.实验结果表明,本方法相比采用单一特征进行识别时,其识别率有显著提升.The extraction and fusion of the features determine the final recognition rate in the research of 3D space handwriting digit. Space handwriting digit features mainly include time-domain features and frequency-domain features. Time-domain fea- tures are typical of acceleration related features, rotation features and short-time energy;frequency-domain features are typical of DCT transform features and FFT transform features. This paper proposes multi-feature fusion method. The various sub-fea- tures were extracted and two sub-features which can reflect characters＇ features were chosen and fused to reduce the dimen- sion of the fused feature vectors. Then the classifier was introduced to categorize and recognize. The experimental results show that the multi-feature fusion method can significantly improve the recognization performance compared with the single teature- based method.

关 键 词：特征融合 旋转特征 DCT变换特征 三维空间手写识别 支持向量机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]