基于二维双向PCA的手写数字识别算法研究  

Research on Handwritten Digit Recognition Algorithm Based on Two Dimension Double PCA

作　　者：郭春妮 高瑜翔[1] 黄坤超[2] GUO Chunni;GAO Yuxiang;HUANG Kunchao(College of Communication Engineering,Chengdu University of Information Technology,Chengdu 610225,China;Southwest China Institute of Electronics Technology,Chengdu 610036,China)

机构地区：[1]成都信息工程大学通信工程学院,四川成都610225 [2]中国西南电子技术研究所,四川成都610036

出　　处：《无线电工程》2019年第12期1099-1102,共4页Radio Engineering

基　　金：省属高校创新团队“新一代无线通信理论与ASIC设计”(15TD0022)

摘　　要：手写数字识别系统中,传统K最近邻(KNN)算法的识别速度较慢,而基于主分量分析(PCA)以及二维主分量分析(2DPCA)的KNN识别算法其识别速度虽有提升,但识别率仍然有待提高。为此,提出了一种基于二维双向主分量分析(Two Dimension Double PCA,2DDPCA)结合KNN的识别算法,并使用MINIST手写数据集进行了仿真实验验证,结果表明,该算法与2DPCA相比,在识别速度相当的情况下,识别率可提高近3个百分点。In the handwritten digit recognition system,the recognition speed of the traditional KNN algorithm is quite slow.Although the recognition speed of the KNN algorithm built on PCA or 2 DPCA is improved,the recognition rate is still not high.Therefore,a recognition algorithm based on two dimension double principal component analysis(2 DDPCA,Two Dimension Double PCA)combined with KNN is proposed,and the simulation experiment and verification are conducted by using MINIST handwritten data set.The results show that compared with 2 DPCA,the recognition rate can be increased by nearly 3 percentage points in the case of a comparable recognition speed.

关 键 词：手写数字识别 K最近邻 二维主分量分析 二维双向主分量分析 识别率 

分 类 号：TN911.73[电子电信—通信与信息系统]