基于ELM和SVM皮电信号情绪分类识别的研究

Classification and Recognition of Electrodermal activity based on ELM and SVM

皮电信号(Electrodermal activity,EDA)是一种不平稳的非周期性微弱信号,能够反映不同情绪状态下人体皮肤内血管的舒张和收缩以及汗腺的分泌活动,在情绪分类识别中具有重要的研究意义。针对EDA信号时域特征提取的分类方法识别率低,极限学习机(Extreme learning machine,ELM)具有训练参数少、学习效率高、泛化能力强等的优点,本研究基于生理信号采集设备采集了12位被试的皮电数据,从时域上进行特征提取,输入KS(Kennard-Stone)模型随机筛选样本,使用支持向量机(Support Vector Machine,SVM)和极限学习机的方法对积极、消极、中性情绪进行分类,并对分类准确率进行比较。实验结果表明,相较于支持向量机分类器55.56%的平均分类准确率,极限学习机平均分类准确率达64.16%,提高了8.6%,采用配对t检验进行验证,t检验结果为P=0.047781453<0.05,具有显著的统计学差异。极限学习机算法适用于情绪分类识别,相较于支持向量机具有更好的情感识别效果。

Electrodermal activity(EDA)is a kind of unstable,non-periodic weak signal,which can reflect the vasodilation and contraction of blood vessels in the skin and the secretion of sweat glands under different emotional states,and has important research significance in the classification and recognition of emotions.The classification method of TIME domain feature extraction of EDA signal has low recognition rate,while Extreme Learning Machine(ELM)has the advantages of few training parameters,high learning efficiency and strong generalization ability.In this study,skin electrical data of 12 subjects were collected based on physiological signal acquisition equipment.The features were extracted from the time domain,and the samples were randomly screened by KS(Kennard-Stone)model.The positive,negative and neutral emotions were classified by Support Vector Machine(SVM)and extreme learning Machine(ELM),and the classification accuracy was compared.The experimental results show that compared with the average classification accuracy of support vector machine classifier of 55.56%,the average classification accuracy of EXTREME learning machine is 64.16%,an increase of 8.6%.Paired T-test is used for verification,and the t-test result isP=0.047781453<0.05,showing a significant statistical difference.The extreme learning machine algorithm is suitable for emotion classification and recognition and has better emotion recognition effect than support vector machine.