| 采用多路图注意力网络的情绪脑电信号识别方法 |
| |
| 引用本文： | 李景聪,潘伟健,林镇远,陈希昶,潘家辉.采用多路图注意力网络的情绪脑电信号识别方法[J].智能系统学报,2022,17(3):531-539. |
| |
| 作者姓名： | 李景聪 潘伟健 林镇远 陈希昶 潘家辉 |
| |
| 作者单位： | 华南师范大学 软件学院，广东 佛山 528200 |
| |
| 摘 要： | 情绪是一种大脑产生的主观认知的概括。脑信号解码技术可以以一种较客观的方式来有效地研究人的情绪及其相关认知行为。本文提出了一种基于图注意力网络的脑电情绪识别方法(multi-path graph attention networks, MPGAT)，该方法通过对脑电信号通道建图，利用卷积层提取脑电信号的时域特征以及各频带的特征，使用图注意力网络进一步捕捉情绪脑电信号的局部特征以及各脑区之间的内在功能关系，进而构建出更好的脑电信号表征。MPGAT在SEED和SEED-IV数据集的跨被试情绪识别平均准确率分别为86.03%、72.71%，在DREAMER数据集的效价(valence)和唤醒(arousal)维度的跨被试平均准确率分别为76.35%和75.46%，达到并部分超过了目前最先进脑电情绪识别方法的性能。本文所提出的脑电信号处理方法有望为情绪认知科学研究与情绪脑机接口系统提供新的技术手段。
|
| 关 键 词： | 情绪识别 跨被试 图卷积神经网路 图注意力网络 脑电信号 脑机接口 神经网络 深度学习 |
Emotional EEG signal recognition method using multi-path graph attention network |
| |
| LI Jingcong,PAN Weijian,LIN Zhenyuan,CHEN Xichang,PAN Jiahui.Emotional EEG signal recognition method using multi-path graph attention network[J].CAAL Transactions on Intelligent Systems,2022,17(3):531-539. |
| |
| Authors: | LI Jingcong PAN Weijian LIN Zhenyuan CHEN Xichang PAN Jiahui |
| |
| Affiliation: | Software College, South China Normal University, Foshan 528200, China |
| |
| Abstract: | Emotion is a generalization of subjective cognition produced by the brain. Brain signal decoding technology can effectively study human emotions and related cognitive behaviors more objectively. This paper proposes a multi-path graph attention networks (MPGAT) method for electroencephalogram (EEG) emotion recognition based on graph attention networks. This method uses convolutional layers to extract the time domain features of EEG signals and the features of each frequency band by mapping EEG signal channels; moreover, it uses the graph attention network to further capture the local features of emotional EEG signals and the internal functional relationships among the brain regions; constructing a better representation of EEG signals later. The average accuracy of MPGAT’s cross-subject emotion recognition in the SEED and SEED-IV datasets were 86.03% and 72.71%, respectively, and the average cross-subject accuracy rates of the valence and arousal dimensions in the DREAMER dataset were 76.35% and 75.46%, respectively, meeting and partially exceeding the performance of the most advanced EEG emotion recognition methods. The EEG signal processing method proposed in this paper can provide novel technical means for the scientific research of emotional cognition and emotional brain-computer interface system. |
| |
| Keywords: | emotion recognition cross-subject graph convolutional neural network graph attention network electroencephalogram signal brain-computer interface neural network deep learning |
|
| 点击此处可从《智能系统学报》浏览原始摘要信息 |
|
点击此处可从《智能系统学报》下载全文 |
