摘要： AdaBoost算法已被广泛地应用于人脸检测系统中，但往往需要大量的训练样本。针对其训练过程复杂冗长的缺陷，选择研究基于少量训练样本的人脸检测问题。采用协方差特征代替图像统计的直方图进行特征提取。为达到更好的分类效果，应用基于Fisher判别式分析的线性超平面分类器，通过AdaBoost算法构成多层级联分类器进行人脸检测。在小数据库里可以看到，与目前用于多数人脸检测系统的类Haar特征相比，该算法在减少训练样本的同时能获得更好的检测效果。

关键词: 人脸检测, 协方差特征, Fisher判别式分析, 训练样本

Abstract: Aiming at the complex and long training process of face detection, this paper addresses the problem of learning to detect faces from a small set of training samples. It proposes to use covariance features to extract the facial features. For better classification performance, linear hyperplane classifier based on Fisher Discriminant Analysis(FDA) is proffered. AdaBoost algorithm is used to construct cascade classifier. It shows that the detection can be significantly improved with the algorithm on a small dataset, compared with Haar-like features used in current most face detection systems.

Key words: face detection, covariance feature, Fisher Discriminant Analysis(FDA), training sample

中图分类号: 

• TP391.41