MLDA+LDA:手写汉字识别中一种两级LDA分类方法

线性判决分析(lineardiscriminateanalysis,LDA)作为一种经典线性工具已经被广泛地运用在各种模式识别问题中,为了降低参数估计误差对于LDA性能的影响,我们提出了一种改进的线性判决分析(modi-fiedlineardiscriminateanalysis,MLDA)算法,并运用到手写汉字识别中,使得识别率有所提高。在此基础上,针对大类别的汉字识别问题,提出了两级LDA的手写汉字识别方法,即MLDA LDA。在对GB2312—80的1034个汉字类别的350套手写样本的实验表明,这个两级LDA识别策略针识别率较最小距离分类器有着3.77%的提高,较LDA 最小距离分类方法有1.71%的提高,表明方法的有效性。

MLDA+LDA:A New Two-stage LDA Method of Handwritten Chinese Character Recognition

As a classical linear tool,Linear Discriminate Analysis (LDA) has been widely used in the field of pattern recognition.For reducing the effect caused by the errors of parameter estimation of LDA.A Modified Linear Discriminate Analysis (MLDA) algorithm,which is used to improve the performance of LDA in Chinese character recognition is proposed.Based on MLDA,this paper also proposes a two-stage LDA recognition scheme for Chinese character recognition,MLDA LDA.Experimental results on 1 034 categories of Chinese character from 350 sets of samples are given.With minimum distance classifier and LDA plus minimum distance classifier,the recognition rate has been increased by 3.77% and 1.71% respectively which shows the efficiency of the proposed approach.