脑-机接口是不依赖外周神经和肌肉的信息传输通道，它为瘫痪病人提供了一种与外界环境通讯的新途径。现有无创脑-机接口所采用的脑电信号和刺激方式尚存在不足，例如闪烁刺激的高对比度引发的视觉疲劳，受试者主动参与的认知脑电没有得到充分利用等，因此开发新的实验范式是一个非常值得研究的问题。本论文设计了一种基于运动起始视觉刺激的脑-机接口实验范式，相比于传统的基于闪烁视觉刺激的脑-机接口，本论文中的范式所使用的视觉刺激具有低亮度、低对比度和非闪烁的优势。首先根据脑电实验研究了所设计范式的诱发响应的生理特征模式，从而为脑-机接口的应用提供生理上的依据和算法设计的参考。使用逐步判别方法对离线数据进行五分类的目标识别达到较高的正确率，证明了所设计脑-机接口范式的可行性。在对实验范式的生理研究的基础上，本文设计实现了一个36选项的字符输入脑-机接口——N200-speller，并研究了N200-speller的诱发响应的时空模式，与P300-speller的生理模式进行比较。结果表明N200-speller诱发响应中最显著的成分是运动起始N2；P300-speller诱发响应中最显著的成分是P300。本文的N200-speller范式的另一个创新点是实验任务的设置，通过受试者的主动认知活动从而提高注意力对诱发响应的增强效应。本论文还研究了基于运动起始视觉刺激范式的脑机接口系统中的模式分类问题。对10位受试者的脑电数据的分类结果表明，使用逐步线性判别、Fisher线性判别和支持向量机算法的分类正确率高于其他方法（近邻法、人工神经网络等）。综上，本论文工作的创新点主要是设计了一种新的基于非闪烁的视觉运动刺激的脑-机接口实验范式，同时引入主动认知任务强化了诱发相应，并在深入研究该范式生理机制的基础上，完成了基于运动起始视觉响应和认知晚成分的脑-机接口系统的构建和可行性分析。
Brain-computer interface (BCI) is a communication channel which does not depend on the brain’s normal output pathways of peripheral nerves and muscles. It provides the paralyzed patients with a novel approach to communicate with the environment. However，the BCI systems using EEG signals usually have the problems such as user’s fatigue caused by uncomfortable flash stimulus (e.g., SSVEP based BCI or P300-speller), lack of employment of BCI user’s active responses.. Consequently, seeking for new BCI paradigm has become a crucial need for in high performance BCI development. Here we proposed a motion-onset paradigm for implementing a new type of BCI, which feautures the advantage of low luminance, low contrast and non-flash. EEG data registered from 15 subjects are used to investigate the spatio-temporal pattern of motion-onset visual evoked potentials (mVEP) in this paradigm. N2 and P2 components, with distinct temporo-occipital and parietal topography, respectively, are selected as the salient features of the brain response to the attended target that the subject selects by gazing at it. The computer determines the attended target by finding which button elicited prominent N2/P2 components. The stepwise linear discriminant analysis is adopted to assess the target detection accuracy of a five-class BCI. The high mean accuracy suggests that the proposed motion-onset BCI could be promising in online implementation.Based on the neurophysiological background of visual motion-onset paradigm, a 36-class BCI named N200-speller is developed in this dissertation. Evoked response of the N200-speller is studied and compared with that of the P300-speller. The comparison indicates that the prominent response of N200-speller is motion-specific N2, while the prominent response of P300-speller is P3. Besides motion stimuli, another feature of the N200-speller is the employment of cognitive experimental task. The user focuses attention on the button labeled with the letter to be communicated and performs color naming task, which helps to enhance the evoked potentials as mVEP and late positive complex. Pattern classification algorithms are applied in the brain-computer interface based on visual motion-onset responses. EEG data sets of ten subjects are analyzed using several algorithms including Stepwise Linear Discriminant Analysis (SWDA), Fisher Linear Discriminant Analysis (FDA), Support Vector Machine (SVM), K-Nearest Neighborhood (KNN), Artificial Neural Network (ANN), etc. Results of classification accuracy indicate that SWDA、FDA and SVM perform better than other methods.In conclusion，a novel BCI based on non-flashing motion-onset paradigm, with cognitive task adopted to enhance the mental responses, was firstly proposed and implemented in this dissertation. This work also studies the neurophysiological patterns of the motion-onset paradigm and the feasibility of this new BCI paradigm.
