摘要： 在很多应用中,组合使用多个分类器可以降低分类错误率。该文就是基于这个思想提出了新的人脸识别算法,即加强概率推理模型。在该算法中,将分类任务划分成多个子分类器,每个子分类器集中于一些难分类的样本,然后组合这些子分类器形成一个强的分类器。试验结果表明算法的识别率比原来的概率推理模型的识别率提高了1.8%。

关键词: 人脸识别, 分类器, 概率推理模型

Abstract: In a wide range of applications,the combination of classifiers leads to substantial reduction of misclassification error.This paper proposes a new algorithm to boost performance of probabilistic reasoning model(PRM) face recognition methods.In this algorithm,it divides the classification task into some sub-classifiers,each sub-classifier concentrates on some labels that are hardest to discriminate.Finally,it combines all these sub-classifiers to form a very strong class ifier experimental result indicates that the proposed methodology enhances performance of the probabilistic reasoning model with an average improvement of correct recognition rate(CRR) up to 1.8%.

中图分类号: 

• TP391.41