摘要： 针对传统主成分分析(PCA)算法提取人脸特征时效率低下的问题,对其求解样本协方差矩阵的特征值和特征向量的过程进行改进,提出一种基于快速PCA降维算法的人脸识别方法。使用交叉验证方法,将支持向量机(SVM)模型训练时的识别准确率作为粒子群的适应度值,利用粒子群优化算法对SVM惩罚参数和核函数参数进行全局寻优,得到参数的全局最优解,用于训练最终的分类器模型进行人脸识别。通过对ORL和Yale数据库中的人脸图像进行实验,结果表明,与传统PCA算法结合SVM模型的识别方法相比,该方法对于人脸图像具有更高的特征提取效率及识别准确率。

关键词: 特征提取, 主成分分析, 粒子群优化, 人脸识别, 支持向量机

Abstract: In order to overcome the low efficiency shortcoming of traditional Principal Component Analysis(PCA) feature extraction,this paper proposes a new face recognition method based on fast PCA dimensionality reduction algorithm which is able to accelerate the process of the eigenvalues and eigenvectors of calculating the sample covariance matrix.In the sense of cross-validation,this paper takes the recognition accuracy of Support Vector Machine(SVM) training model as the fitness value of the Particle Swarm Optimization(PSO),searches globally for the optimal values of penalty parameter and kernel function parameter of SVM,obtains the global optimal values of the parameters,and takes the values to train a final classifier model.Experimental results on the face images of ORL and Yale library show that the new method has higher feature extraction efficiency and recognition accuracy compared with the recognition method based on the traditional PCA and SVM algorithm.

Key words: feature extraction, Principal Component Analysis(PCA), Particle Swarm Optimization(PSO), face recognition, Support Vector Machine(SVM)

中图分类号: 

• TP391.41