Abstract: In this paper，novel bilateral two-dimensional weighted principle component analysis is applied on facial feature extraction.Quantity of feature what extracted is small because image matrix’s dimension is reduced from plane and upright orientation.Considering the fact that the different face regions contain the different recognition information，and the different face regions are given the different weighted values.Experimental result shows this method is quicker and improves recognition rate than two-dimensional principle component analysis.

Key words: facial expression recognition, feature extraction, two-dimension principle component analysis, bilateral two-dimension weighted principle component analysis

摘要： 提出了一种双向二维加权主元分析方法用于人脸表情特征提取，该方法从水平和垂直两个方向对图像矩阵进行降维处理，大幅降低了所提取的特征数目；且考虑到人脸不同部位包含不同的表情信息这一特点，对各个特征赋予不同的权重系数。实验证明，与已有的二维主元分析相比较，该方法不但运算速度快，且获得了更高的识别率。

关键词: 表情识别, 特征提取, 二维主元分析, 双向二维加权主元分析