面向情绪脑电分析的增强型功率谱密度特征提取方法

Feature extraction method based on enhanced power spectral density for emotion analysis using EEG

针对功率谱密度在脑电情绪分析中存在特征单一且无法有效表示频率间差异的问题,提出一种增强型功率谱密度特征提取方法,实现对情绪的分析与差异显著性判断。该方法通过脑电信号的α频率功率谱密度得到功率谱密度图像,利用图像特征提取算法提取其颜色特征、纹理特征与相似性特征,再基于相关性准则剔除冗余特征,以差异显著性P值的最小平均值为目标,获得最终的特征子集,从而有效地融合了不同图像特征,最后对被试的情绪进行分析与差异显著性判断。试验结果表明,所提出的方法能够有效量化SEED数据集中被试的情绪差异;在自行设计情绪脑电试验中,与其他方法相比,利用所提出的方法得到的差异显著性值更小,证明了方法的可行性和有效性。

Aiming at the problem that power spectral density(PSD) has single feature and cannot effectively represent the differences between frequencies in electroencephalogram(EEG)-based emotion analysis,a feature extraction method based on enhanced PSD is proposed to realize the analysis on emotions and the assessment of the significance of difference.After obtaining power spectral density image by α frequency power spectral density of EEG signal,the color feature,texture feature and similarity feature are extracted by image feature extraction algorithm.Then the redundant features are eliminated based on correlation criterion,and the final feature subset is obtained by taking the minimum average value of the significance of difference(P value)as the target,thus effectively fusing different image features.Finally,the emotions of subjects are analyzed,and the significance of the difference is assessed.The experimental results show that the proposed method can effectively quantify the emotional differences of the subjects in the SEED dataset.In the self-designed emotional EEG test,the significance of difference obtained by the proposed method is smaller than other methods,which proves the feasibility and effectiveness of the method.