Abstract:

A new method based on the combination of?Rapid Serial Visual Presentation（RSVP） and Steady-State Visually Evoked Potential（SSVEP） for Electroencephalogram（EEG） stimulation and multi-class events detection is presented in this paper. The artifacts and noise of the raw EEG signals are removed by preprocessing including potential re-reference, baseline removal and spatial filtering. Machine learning algorithms such as Bagging Tree（BT） and Supported Vector Machine（SVM） are used to detect multi-class events of EEG signals induced by double stimuli on 14 subjects. The experimental results show that the EEG signals induced by the combined stimulation have good multi-class separability, which provides a new and effective way to develop the hybrid brain-computer interface application based on RSVP and SSVEP combination paradigm. The outcome also proves that the performance of the BT and SVM models for multi-class recognition of EEG signals induced by RSVP and SSVEP is obviously better than that of the Traditional Canonical Correlation analysis（CCA） algorithm.

Key words: Electroencephalogram（EEG）, Rapid Serial Visual Presentation（RSVP）, Steady-State Visually Evoked Potential（SSVEP）, Bagging Tree（BT）, Supported Vector Machine（SVM）, multi-class detection

摘要：

提出一种新的基于快速序列视觉呈现（Rapid Serial Visual Presentation，RSVP）与稳态视觉诱发电位（Steady-State Visually Evoked Potential，SSVEP）组合范式的脑电信号（Electroencephalogram，EEG）刺激与多类事件检测方法。对诱发的原始脑电信号通过电位重参考、基线去除、空间滤波等预处理操作去除数据的伪迹和噪声，通过自举聚合决策树（Bagging Tree，BT）和支持向量机（Supported Vector Machine，SVM）等机器学习算法，对14名受试者双重刺激诱发的脑电信号进行目标与频率相结合的多类事件检测，通过实验验证了该组合范式诱发的脑电信号具有良好的多类可分性，为开发基于RSVP和SSVEP两种范式的混合型脑-机接口应用提供了一种新的有效途径。同时，实验结果还表明，基于机器学习的BT和SVM模型对RSVP和SSVEP组合范式诱发的EEG信号进行多类识别的性能明显优于传统的典型关联分析（Canonical Correlation Analysis，CCA）算法的性能。

关键词: 脑电信号, 快速序列视觉呈现, 稳态视觉诱发, 决策树, 支持向量机, 多类检测