基于时频域组合特征的脑电信号情感分类算法

Emotional Classification Algorithm of Electroencephalogram Signals Based on Time-frequency Combination Characteristics

为了提高基于脑电信号(electroencephalogram,EEG)情感识别的准确率,提取了脑电信号的时域与频域特征,并且将其进行组合形成时频域组合特征,作为不同识别模型下的输入。采用集成决策树(bagging tree,BT)、贝叶斯线性分析(Bayesian linear discriminant analysis,BLDA)、线性判别分析(linear discriminant analysis,LDA)及支持向量机(support vector machine,SVM)四种浅层机器学习算法对EEG在效价与唤醒度上进行二分类情感识别。实验结果表明,DEAP数据集在效价上,基于时频域组合特征在BT分类器下的识别精度平均达到92.54%,在唤醒度维度上基于时频域组合特征在SVM下平均识别精度达到94.62%。

In order to improve the accuracy of emotion recognition based on electroencephalogram(EEG),temporal and frequency-domain features of EEG signals are extracted and combined to form a temporal and frequency-domain signature,which can be input in different recognition models.Four kinds of shallow machine learning algorithms,namely,Bagging tree(BT),Bayesian linear discriminant analysis(BLDA),linear discriminant analysis(LDA)and support vector machine(SVM),were applied to classify EEG emotion on valency and arousal.The results show that in terms of titer,the recognition accuracy of DEAP data set based on time-frequency domain combination features under BT classifier is 92.54% on average,and that based on time-frequency domain combination features in SVM is 94.62% on average in terms of arousal degree.