用基于平行坐标图的可视化人机交互技术提取EEG信号特征的方法

Feature extraction of EEG signals using the visualized BCI technique based on parallel coordinates plot

应用可视化人机交互(HCI)方法进行了脑电图记录(EEG)信号特征提取技术的研究。该研究一方面在脑机接口(BCI)技术领域提出了一种新的特征提取技术方法,同时通过可视化人机交互的专家智慧参与,实现了面向对象领域和面向数据模式识别的有效结合,克服了单一机器学习的局限性。首先介绍了多元图表示的基本理论,然后提出了基于平行坐标图的可视化人机交互技术,接着进行了单通道和多通道EEG信号特征提取的可视化人机交互技术的研究,最后采用第二届国际脑机接口竞赛中的数据集Ⅳ进行了数据实验。实验表明,本文提出的方法的识别结果优于实验数据集国际竞赛最优结果和文献报道中的当前国际最优结果。

The study on the techniques for extraction of electroencephalography （EEG） singnal features was conducted based on human-computer interaction （HCI）. It proposes a new way of feature extraction in the technical field of brain-computer interface （BCI）, and through the expert intelligence endeavor to feature extraction by the HCI based on graphical presentation of multivariate data, effectively realized the combination of data-oriented pattern recognition and object-oriented domains, and overcame the obstacles of the only mechine learning. It firstly introduced the visualized HCI technique based on graphical presentation of multivariate data, then studied the visualized feature extraction techniques for single channel and multi-channel EEG signals. The experiments were performed based on the dataset IV of the international BCI competition II. The experimental results were very superior to that of the intemational BCI competition II and the previously reported optimal classification performance of the international compared baseline methods. It proved the validity of the research methods in this paper.