小样本条件下的人脸特征提取算法

传统的“特征脸”和“Fisher脸”对于小样本的识别非常不理想。提出了基于小波变换的梯度方向直方图结合线性判别分析的特征提取算法，对原图像进行小波变换，对近似部分进行HOG+LDA处理，对高频部分用均方差处理，分别对这两部分特征进行欧氏距离判断，进行加权识别。实验数据表明该方法能够克服LDA对小样本敏感问题，提高识别率的同时降低了耗时。

Face feature extraction algorithm for small training sets

Wavelet transforms-based Histogram of Oriented Gradient（HOG） with Linear Discriminant Analysis（LDA） for feature extraction is proposed since Eigenface and Fisherface do not work well to cope with small training sets of high dimension. In order to represent face features better and reduce dimension, wavelet transform could be used for extracting face features. The approximation coefficients are processed by HOG+LDA technique and mean square deviation is employed to handle horizontal, vertical and diagonal detail coefficients, respectively. The resulting features are discriminated by Euclidean distance. Simulation results conducted on the ORL and Yale databases show that the proposed method achieves excellent performance both in terms of classification accuracy and computational efficiency.