计算机工程与应用 ›› 2023, Vol. 59 ›› Issue (11): 160-167.DOI: 10.3778/j.issn.1002-8331.2202-0311

• 模式识别与人工智能 • 上一篇    下一篇

基于图表征和双重注意力机制的跨被试ERP检测

相晓嘉,兰珍,闫超,李子杏,唐邓清,周晗   

  1. 国防科技大学 智能科学学院,长沙 410073
  • 出版日期:2023-06-01 发布日期:2023-06-01

Cross-Subject ERP Detection Based on Graph and Dual Attention Mechanism

XIANG Xiaojia, LAN Zhen, YAN Chao, LI Zixing, TANG Dengqing, ZHOU Han   

  1. College of Intelligence Science and Technology, National University of Defense Technology, Changsha 410073, China
  • Online:2023-06-01 Published:2023-06-01

摘要: 针对事件相关电位(event-related potential,ERP)在跨被试场景下检测精度不高的问题,提出了一种基于图表征和双重注意力机制的卷积循环神经网络模型。该模型采用不依赖于被试和任务的图来表征脑电信号中的空间信息,并级联卷积神经网络(convolutional neural network,CNN)和长短期记忆网络(long short-term memory network,LSTM)形成CNN-LSTM基础框架,同时嵌入双重注意力机制(即选择性内核卷积和自注意力机制)以充分提取不同被试脑电信号的时空特征,从而提高跨被试场景下的ERP检测精度。在基于快速序列视觉呈现范式的大规模基准数据集上的实验结果表明,与现有的7种ERP检测方法相比,所提方法在跨被试场景下具有显著的优越性。

关键词: 脑电图(EEG), 相关电位(ERP)检测, 图表征, 注意力机制, 卷积循环模型

Abstract: In order to improve the detection accuracy of event-related potential(ERP) in subject-independent scenarios, a convolutional recurrent neural network model based on graph embedding and dual attention mechanisms is proposed. The model uses a graph to represent the spatial information in electroencephalogram(EEG) signals, and uses the cascade framework of convolutional neural network(CNN) and long short-term memory network(LSTM) as the basic framework. By embedding dual attention mechanisms(i.e., selective kernel convolution and self-attention mechanism), it can fully extract the temporal and spatial features of EEG signals of different subjects, so as to improve the ERP detection accuracy in subject-independent scenarios. A large number of experiments carried out on the benchmark dataset based on rapid serial visual presentation paradigm demonstrate that the proposed method has significant superiority over 7 existing ERP detection methods in subject-independent scenarios.

Key words: electroencephalogram(EEG), event-related potential(ERP) detection, graph, attention mechanism, convolution recurrent model