Abstract: A facial expression recognition method based on double discriminant subspace and radial basis function neural network(RBFNN) is proposed in this paper.First,this method apply the KPCA and FLD algorithm to carry out discriminant analysis in double discriminant subspace(null space and range space) and extract two kinds of expression feature discriminant information:regular information and irregular information,and fuse them by certain rule;Then the RBFNN and the fusion feature information can be use to facial expression recognition. Experimental result on JAFFE show that it is a valid method for facial expression recognition.

Key words: double discriminant subspace, RBFNN, facial expression recognition, KPCA, FLD

摘要： 提出了一种新的基于双决策子空间和径向基函数（RBF）神经网络的人脸表情识别方法。该方法首先采用KPCA+ FLD算法在双决策子空间（核空间和值域空间）中进行决策分析，提取两类判决特征信息：非常规信息和常规信息，并按一定的规则融合这两类判决信息；再运用RBF神经网络分类器和融合特征信息进行人脸表情的分类识别。基于日本女性表情数据库JAFFE的实验结果表明，它是一种有效的人脸表情识别方法。

关键词: 双决策子空间, RBF神经网络, 人脸表情识别, 核主元分析, Fisher线性判别分析