Abstract: As most of the face recognition techniques will suffer serious performance drop when there is only one training sample per person, a face recognition algorithm based on sample augment methods and improved Two Dimensional Principal Component Analysis (2DPCA) was proposed. By analyzing the advantage and disadvantage of various sample augment methods, some of them were combined to synthesize virtual samples in order to make full use of the single training image. Improved 2DPCA was chosen to extract the feature of the synthetic virtual samples. The training samples were divided into sub-blocks and then the covariance matrix was constructed by these sub-blocks which were normalized by the within-class average value in each sub-block. The experimental results on ORL and Yale face database indicate that the performance of the proposed algorithm is better than those of general 2DPCA and the method only using sample augment.

Key words: single sample, face recognition, sample augment, within-class average value, Two Dimensional Principal Component Analysis (2DPCA)

摘要： 针对大多数人脸识别方法在单个训练样本条件下识别性能下降的问题,提出了结合多种样本扩充方法和改进二维主成分分析(2DPCA)的人脸识别算法。通过分析各种样本扩充方法的优缺点,用多种样本扩充方法来生成虚拟样本,以充分利用单一样本所提供的信息。采用改进的2DPCA方法对生成的虚拟样本进行特征提取,对训练样本进行分块,并用类内平均值规范后的分块来构造总体散布矩阵。在ORL和Yale人脸库上的实验表明,所提出的方法在识别性能方面优于普通的2DPCA方法,优于单一的样本扩充方法。

关键词: 单样本, 人脸识别, 样本扩充, 类内平均值, 二维主成分分析(2DPCA)