基于PCA-SVM多生理信息融合的情绪识别方法  

An emotion recognition method of multiphysiological information fusion based on PCA-SVM

作　　者：李发权[1] 杨立才[1] 颜红博 

机构地区：[1]山东大学控制科学与工程学院,山东济南250061

出　　处：《山东大学学报（工学版）》2014年第6期70-76,共7页Journal of Shandong University（Engineering Science）

摘　　要：为了有效解决情绪识别过程中多种生理信息融合所导致的运算量过大的问题,提出了一种主成分分析(principal component analysis,PCA)与支持向量机(support vector machine,SVM)相结合的情绪识别方法。利用主成分分析法,求出各特征对情绪识别效果的影响权重,通过阈值法选择权重较大的特征组成新的特征子集,从而减少SVM的输入特征维数,降低算法的运算量。试验结果表明,该方法可以有效提高算法的执行效率。To reduce the complexity of the emotion-recognition algorithm caused by multiphysiological information fu-sion an emotion recognition method based on Principal Component Analysis （PCA ） and Support Vector Machine （SVM）was proposed.The influential weights of emotion recognition were calculated for initial features by the PCA, and the features of which the weights were larger than a certain threshold were selected to compose the new feature set. Thus the dimension of the classifierinputs could be reduced so that the complexity of the algorithm will be simplified. Experimental results showed that the PCA-SVM algorithm for sentiment analysis could effectively improve the efficiency of emotion recognition.

关 键 词：主成分分析 支持向量机 信息融合 情绪识别 特征子集 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]