基于PCA和LDA融合算法的性别鉴别

Sex identification algorithm based on PCA and LDA

结合主元分析(PCA)与线性鉴别分析(LDA)的特点,利用PCA-LDA算法进行性别鉴别。通过PCA算法求得训练样本的特征子空间,并在此基础上计算出LDA算法的特征子空间。将PCA算法与LDA算法的特征子空间进行融合,获得PCA-LDA算法的融合特征空间。训练样本与测试样本分别朝融合特征空间投影,从而得到识别特征。利用最近邻准则即可完成性别鉴别。实验中利用三种预处理方法(PCA+LDA、HG+PCA+LDA、RHG+PCA+LDA),得出各自的实验结果,并进行比较。实验结果表明,利用RHG+PCA+LDA方法预处理后,使用PCA-LDA算法进行性别鉴别可以得到理想的效果。

Combined with the advantages of principal components analysis （PCA） andlineardiscriminantanalysis（LDA）,PCA-LDA on sex classification was presented. Feature sub-space of training samples was obtained by the way of PCA, and feature sub-space from LDA was calculated on the basis of PCA.Meanwhile, the two feature sub-spaces from PCA and LDA were fused, and the fusion feature space was acquired.After trainingsamples and test samples were respectively projected towards the fusion feature space, recognition features were accordingly gained. Nearest neighborrule was utilized in sex classification. There were three kinds of pretreatment methods in the experiment, PCA＋LDA, HG＋PCA＋LDA and RHG＋PCA＋LDA, the respective results were presented in the article, and also some comparisons had been made out. The experiment results show that with RHG＋PCA＋LDA as the pretreatment method, the sex identification can receive desired results with the algorithm of PCA＋LDA.