摘要
对线性回归分类算法进行了改进。考虑了线性回归分类算法中没有考虑的类间信息,通过选择类模式的投影方向判别不同类的模式,不同类的模式互相远离,相同类的模式尽可能靠近来估计投影矩阵;再利用投影矩阵将训练图像及测试图像投影到各类的特征子空间;最后,计算出测试图像与训练图像间的距离,利用K-近邻分类器完成人脸的识别。在FERET人脸数据库上进行实验验证。实验结果表明,相比其他回归分类算法,本算法取得了更好的识别效果。
Linear regression classification algorithm is improved . Consider the information between the classes that linear regression classification algorithm does not be considered . The modes of different class are identified through choosing the projection directions of class mode .Away from each other different types of modes ,the modes of the same class as close as possible to estimate the projection matrix ;Then the training images and the various types of test image is projected onto the subspace by using projection matrix ;Finally ,K-nearest neighbor classifier is used to finish face recognition after calculating the distance between the test image and the training images .Conduct experiments on FERET face database .Experimental results show that compared to other regression classification algorithm ,the algorithm achieved a better recognition effect .
出处
《皖西学院学报》
2014年第5期28-30,共3页
Journal of West Anhui University
基金
甘肃省教育厅科研项目(2013A-124)
甘肃省自然科学基金资助项目(1107RJZA170)
关键词
人脸识别
线性判别
线性回归分类
K-近邻分类器
face recognition
linear discriminant
linear regression classification
K-nearest neighbor classifier
