【摘要】 提出了一种旨在减少支持向量机的训练量和提高特征有效性的表情识别算法。使用排序PCA+LDA得到最优表情向量;使用模糊核聚类进行有效数据集约简,构建二叉决策树训练支持向量机。在JAFFE数据库上的识别结果优于其它几种算法,在保证识别率的同时缩短了训练时间。更多还原

【Abstract】 To describe an expression recognition algrithm aims at reducing the training volume and enhancing the validity of features.It employed the sort PCA and LDA to get the optimal expression vector,and used FKC to reduce effective dataset again and build binary de-cision tree to train SVM.Experimental results in the JAFFE database indicate that the proposed algorithm generates higher accuracy than others and shorten the training time simultaneity.更多还原