论文部分内容阅读
本研究针对脑电信号在采集过程中出现的漂移情况,采用支持向量机分类器,分析了节律对数功率、分形维数和信号熵等9种特征,研究了脑电漂移数据对情绪分类的影响;同时,采用拟合求差的方法,尝试对脑电漂移数据进行校正.实验结果表明:脑电漂移数据会导致情绪分类正确率下降,而拟合求差法可以在一定程度上补偿漂移数据对分类造成的不利影响.仿真结果显示:不存在漂移数据时,样本熵和θ节律功率对数两种特征的情绪分类效果最好,而存在未经校正的漂移数据时,δ节律功率对数特征的情绪分类结果最好;漂移数据校正后,样本熵和δ节律功率对数两种特征的情绪分类结果最好.
In this study, we focused on the drift of EEG signals in the process of acquisition, and used support vector machine (SVM) classifier to analyze nine characteristics of rhythmic logarithm power, fractal dimension and signal entropy. The effects of EEG drift data on emotion classification At the same time, we try to correct the data of EEG drift by using the method of fitting difference.Experimental results show that the EEG drift data will lead to a decrease in the accuracy rate of emotion classification, and the fitting difference method can compensate drift to a certain extent And the adverse effect of data on the classification.The simulation results show that the emotion classification of sample entropy and θ rhythm power logarithm is the best in the absence of drift data and the δ rhythmicity power paired with uncorrected drift data The emotion classification results of the number features are the best. After the drift data are corrected, the emotion classification results of the sample entropy and the logarithm of the rhythmicity of the rhythm are the best.