Abstract: In view of the disadvantages that Isometric mapping （Isomap） algorithm is unable to follow-up the test samples which are collected separately and to use dimensionality reduction, and Isomap algorithm does not make use of the classification of information sample point as well, a method was proposed to recognize the face using supervised incremental isometric mapping algorithm (SIIsomap), and combined wavelet transformation to pretreat the images. The experimental results based on ORL database show that SIIsomap algorithm compared with Isomap algorithm would greatly reduce calculation time in handling additional samples and improve recognition accuracy.

Key words: dimensionality reduction, face recognition, wavelet transform, isometric mapping

摘要： 针对等距离映射（Isomap）算法无法对后续采集的测试样本单独进行降维处理和未能利用样本点分类信息的不足，提出了一种有监督的增量式等距离映射算法（SIIsomap），并采取小波变换对图像进行预处理。通过对ORL数据库实验证明，SIIsomap算法与Isomap算法相比大大降低了处理新增样本点的计算时间，并且提高了识别精度。

关键词: 数据降维, 人脸识别, 小波变换, 等距离映射