Abstract: In order to resolve the problem that the feature vector dimension is too high，rough set attribute reduction algorithm applied in expression features selection is advanced.It can acquire local expression features vectors using geometrical feature points method.Rough set theory is introduced，and the local expression features are optimized and selected with an attribute reduction algorithm of rough set.Redundant features and irrelated information to expression classification are also eliminated.Experimental results show that this method can be realized simply.High recognition rate and speed also indicate the effect of this facial expression features selection.

Key words: rough set, features selection, geometrical feature points, expression classification

摘要： 为解决取得特征向量维数过高问题，提出了一种改进的粗糙集属性约简算法。运用几何特征点方法得到人脸表情的局部特征向量，引入粗糙集理论，用改进的属性约简算法对提取到的表情特征进行优化选择，去掉冗余特征和对表情分类无用的不相关信息。实验结果显示，该方法不仅实现方便，识别率高，识别所用的时间也大大减少，充分表明了该方法的有效性。

关键词: 粗糙集, 特征选择, 几何特征点, 表情分类