多预测融合的脑电情绪识别迁移方法

Transfer Method for EEG Emotion Recognition Based on Multi-Prediction Fusion

针对脑电情绪识别中脑电数据的概率分布会因个体和时间而发生变化的问题,提出一个称为领域对抗多预测融合(Domain-Adversarial Multi-Prediction Fusion,DAMPF)的迁移方法。无需目标域数据参与模型训练,该方法通过领域对抗训练来增强所提取特征的可迁移性,并为每个源域学习一个独立的情绪标签预测器。目标域中每个样本的情绪类别均由一个将所有标签预测器的预测进行平均的融合策略来决定。实验结果表明,所提方法优于传统的领域自适应方法和仅有单一预测器的领域泛化方法,验证了所提方法在脑电情绪识别迁移问题中的有效性。

In order to solve the problem that the probability distribution of electroencephalography(EEG)data in EEG emotion recognition varies across individuals and over time,this paper proposes a transfer method called Domain-Adversarial Multi-Prediction Fusion(DAMPF).Without the need for target domain data to participate in model training,the method enhances the transferability of the extracted features through domain-adversarial training and learns an independent emotional label predictor for each source domain.The emotion category of each sample in the target domain is determined by a fusion strategy that averages the predictions of all label predictors.The experimental results show that the proposed method outperforms the traditional domain adaptation methods and the domain generalization method with only a single predictor,which verifies the effectiveness of the proposed method in the transfer problem of EEG emotion recognition.