基于IMF能量熵的脑电情感特征提取研究
发布时间：2019-02-12 09:58
【摘要】：为提高脑电信号情感识别分类准确率,结合经验模态(EMD)分解和能量熵提出一种新的脑电特征提取方法。本研究主要介绍了EMD分解的基本原理,分析了传统EMD算法中的"端点效应",采用分段幂函数插值算法改善了EMD分解的精度和性能,然后将改进后的算法应用到脑电信号特征提取,获取脑电信号的IMF分量后计算出IMF能量熵作为情感识别的特征,最后通过分类实验对比改进后的EMD算法和传统EMD算法对脑电情感特征的分类准确率。实验结果显示改进的EMD算法能使识别率提高15%左右,并且以IMF能量熵为特征的平均识别率在80%以上,实验结果表明将IMF能量熵用于脑电信号情感识别是可行的。
[Abstract]:In order to improve the classification accuracy of EEG emotion recognition, a new EEG feature extraction method combining empirical mode (EMD) decomposition and energy entropy is proposed. This paper mainly introduces the basic principle of EMD decomposition, analyzes the "endpoint effect" in traditional EMD algorithm, and improves the precision and performance of EMD decomposition by using piecewise power function interpolation algorithm. Then the improved algorithm is applied to the feature extraction of EEG signal, and then the IMF energy entropy is calculated as the feature of emotion recognition by obtaining the IMF component of EEG signal. Finally, the classification accuracy of the improved EMD algorithm and the traditional EMD algorithm is compared with the classification experiment. The experimental results show that the improved EMD algorithm can increase the recognition rate by about 15%, and the average recognition rate based on the IMF energy entropy is over 80%. The experimental results show that it is feasible to apply the IMF energy entropy to EEG emotion recognition.
【作者单位】： 华东理工大学;
【基金】：国家自然科学基金资助项目(61071085) 上海市科委科技创新行动计划生物医药领域产学研合作项目(12DZ1940903)
【分类号】：TN911.7;R338
，
本文编号：2420332
[Abstract]:In order to improve the classification accuracy of EEG emotion recognition, a new EEG feature extraction method combining empirical mode (EMD) decomposition and energy entropy is proposed. This paper mainly introduces the basic principle of EMD decomposition, analyzes the "endpoint effect" in traditional EMD algorithm, and improves the precision and performance of EMD decomposition by using piecewise power function interpolation algorithm. Then the improved algorithm is applied to the feature extraction of EEG signal, and then the IMF energy entropy is calculated as the feature of emotion recognition by obtaining the IMF component of EEG signal. Finally, the classification accuracy of the improved EMD algorithm and the traditional EMD algorithm is compared with the classification experiment. The experimental results show that the improved EMD algorithm can increase the recognition rate by about 15%, and the average recognition rate based on the IMF energy entropy is over 80%. The experimental results show that it is feasible to apply the IMF energy entropy to EEG emotion recognition.
【作者单位】： 华东理工大学;
【基金】：国家自然科学基金资助项目(61071085) 上海市科委科技创新行动计划生物医药领域产学研合作项目(12DZ1940903)
【分类号】：TN911.7;R338
，
本文编号：2420332
本文链接：https://www.wllwen.com/xiyixuelunwen/2420332.html
最近更新
教材专著
