摘要： 根据隐马尔可夫模型(HMM)适用于处理连续动态序列信号、支持向量机(SVM)与K近邻分类器(KNN)擅长模式分类的特点，设计一种(HMM+KNN)+SVM的混合分类器。利用HMM与KNN对测试样本进行判决。当判决结果相同时，直接输出判决结果，否则引入SVM对测试样本进行再判决。实验结果表明，该方法所确定的分类器优于单一的分类器判决，能有效实现表情识别。

关键词: 表情识别, 隐马尔可夫模型, 支持向量机, K近邻距离分类器

Abstract: According to the function of Hidden Markov Model(HMM) in processing continuous dynamic signal and model recognition, Support Vector Machine(SVM) and KNearest Neighbor classifier(KNN) being adapt to deal with pattern classification, this paper presents a new facial expression recognition model based on (HMM+KNN)+SVM. It gets training samples feature sequences, and constructs HMM and SVM, at the same time, training samples are set as KNN classifier’s template. It gets the class results of test sample by HMM and KNN respectively, if these two results are the same, then puts out the result as the test sample’s decision class, otherwise, calls SVM to get the test sample’s class result. Experimental result shows that the method is better than single classifier.

Key words: facial expression recognition, Hidden Markov Model(HMM), Support Vector Machine(SVM), KNearest Neighbor classifier(KNN)

中图分类号: 

• TP311.52