小波变换结合经验模态分解在音乐干预脑电分析中的应用
发布时间：2018-08-21 10:38
【摘要】：本文旨在结合小波分析与经验模态分解(EMD),充分提取音乐干预下的脑电(EEG)信号特征参数,提高情绪状态评估的分类准确率与可靠性,以期为辅助音乐治疗提供支持与帮助。采用音乐诱发情绪的多通道标准情感数据库(DEAP)中的数据,利用小波变换提取出额区(F3,F4)、颞区(T7,T8)和中央(C3,C4)通道的α波、β波以及θ波节律;对提取的脑电节律进行EMD以获得固有模态函数(IMF)分量,再进一步提取脑电节律波的IMF分量平均能量和幅度差特征值,即每种节律波中包含3个平均能量特征和2个幅度差特征值,以达到充分提取EEG特征信息的目的;最后基于支持向量机分类器实现情感状态评估。结果表明,利用该算法可以使无情绪、积极情绪、消极情绪之间分类最优正确率达到100%,使得积极与消极情绪之间的识别率提升10%左右,可以实现无情绪与积极、无情绪与消极情绪等情感状态的有效评估。处于不同情感状态下,音乐治疗效果差异较大,提高情感状态评估的分类正确率,将帮助提高音乐治疗的效果,更好地为音乐治疗提供支持。
[Abstract]:In this paper, wavelet analysis and empirical mode decomposition (EMD),) are used to fully extract the characteristic parameters of EEG (EEG) signal under music intervention, and to improve the classification accuracy and reliability of emotional state evaluation, in order to provide support and help for music therapy. The rhythm of 伪 wave, 尾 wave and 胃 wave of frontal region (F3F4), temporal region (T7T8) and central (C3OC4) channel were extracted by wavelet transform from the multichannel standard emotion database (DEAP). The extracted EEG rhythm is extracted by EMD to obtain the (IMF) component of the intrinsic mode function, and the eigenvalues of the average energy and amplitude difference of the IMF component of the EEG rhythm wave are further extracted. That is, each rhythm wave contains three average energy features and two amplitude difference eigenvalues to fully extract EEG feature information. Finally, emotion state evaluation is realized based on support vector machine classifier. The results show that this algorithm can make the optimal classification accuracy of non-emotion, positive emotion and negative emotion up to 100, and the recognition rate between positive and negative emotions can be increased by about 10%. Effective assessment of emotional states such as no emotion and negative emotion. In different emotional states, the effect of music therapy is quite different. Improving the classification accuracy of emotional state evaluation will help to improve the effect of music therapy and provide better support for music therapy.
【作者单位】： 燕山大学电气工程学院生物医学工程研究所;河北省测试计量技术及仪器重点实验室;北京工业大学生命科学与生物工程学院;前景光电技术有限公司;
【基金】：河北省自然科学基金资助项目(F2014203244) 中国博士后科学基金资助项目(2014M550582)
【分类号】：R493;TN911.7
，
本文编号：2195441
[Abstract]:In this paper, wavelet analysis and empirical mode decomposition (EMD),) are used to fully extract the characteristic parameters of EEG (EEG) signal under music intervention, and to improve the classification accuracy and reliability of emotional state evaluation, in order to provide support and help for music therapy. The rhythm of 伪 wave, 尾 wave and 胃 wave of frontal region (F3F4), temporal region (T7T8) and central (C3OC4) channel were extracted by wavelet transform from the multichannel standard emotion database (DEAP). The extracted EEG rhythm is extracted by EMD to obtain the (IMF) component of the intrinsic mode function, and the eigenvalues of the average energy and amplitude difference of the IMF component of the EEG rhythm wave are further extracted. That is, each rhythm wave contains three average energy features and two amplitude difference eigenvalues to fully extract EEG feature information. Finally, emotion state evaluation is realized based on support vector machine classifier. The results show that this algorithm can make the optimal classification accuracy of non-emotion, positive emotion and negative emotion up to 100, and the recognition rate between positive and negative emotions can be increased by about 10%. Effective assessment of emotional states such as no emotion and negative emotion. In different emotional states, the effect of music therapy is quite different. Improving the classification accuracy of emotional state evaluation will help to improve the effect of music therapy and provide better support for music therapy.
【作者单位】： 燕山大学电气工程学院生物医学工程研究所;河北省测试计量技术及仪器重点实验室;北京工业大学生命科学与生物工程学院;前景光电技术有限公司;
【基金】：河北省自然科学基金资助项目(F2014203244) 中国博士后科学基金资助项目(2014M550582)
【分类号】：R493;TN911.7
，
本文编号：2195441
本文链接：https://www.wllwen.com/huliyixuelunwen/2195441.html
最近更新
教材专著
