基于事件相关电位的脑机接口系统研究与设计

发布时间：2018-07-18 15:52

【摘要】：脑-机接口(brain-computer interface, BCI)是一种新的信息交流系统，，不需要语言或者动作，就可以在人脑和计算机之间进行直接的交流和控制。脑机接口技术涉及到神经学、心理认知科学、康复工程、生物医学工程和计算机科学等多个学科的交叉。可以通过脑电图（EEG）、皮层脑电图（EcoG）、功能性磁共振成像（fMRI）、功能性近红外光学成像（fNIR）等方法来对BCI进行研究。在对近年来国内外的BCI系统进行分析之后，本文对基于事件相关电位（ERP）的脑机接口系统进行了研究。采用无创伤的基于EEG的BCI研究方法，设计了基于事件相关电位的新型输入系统，可以直接通过脑电信号而不用手就能进行密码的输入。本文研究的内容包括输入系统实验范式的设计，系统整体的设计与实现，脑电信号的分析与处理等。在设计实验范式时，针对ERP中P300的电位特性，分别对输入字符的形式及颜色进行了对比，并在此基础上采用不同的实验参数进行了多次实验。本文在根据实验范式进行系统实现时，主要利用VC++中的MFC来进行编程。该系统集刺激界面的产生、脑电信号采集、存储及分析于一体，可以实时的将被试者所注视的目标字符输出到显示界面上。本文在对第三届BCI竞赛中P300数据进行分析的基础上，确定出对自采集脑电信号的处理方案。对于竞赛数据，首先利用数字滤波器进行滤波，采用降采样及时间窗进行降维，然后利用Fisher线性判别、正则化Fisher、核Fisher、支持向量机、贝叶斯线性判别进行分类。结果表面正则化Fisher在较短的时间内使用较少的叠加次数就达到最高的正确率，Fisher线性判别所达到的效果次之。所以，在对自采集数据使用叠加平均、独立成份分析（ICA）等方法去除干扰之后，采用Fisher线性判别及正则化Fisher进行分类。最终确定出最佳的输入方式、实验参数及脑电信号处理方案。
[Abstract]:Brain-computer interface (BCI) is a new information exchange system, which can communicate and control directly between human brain and computer without the need of language or action. Brain-computer interface technology involves the intersection of many subjects, such as neurology, psychological cognitive science, rehabilitation engineering, biomedical engineering and computer science. BCI can be studied by electroencephalogram (EEG), cortical electroencephalogram (EcoG), functional magnetic resonance imaging (fMRI) and functional near infrared optical imaging (fNIR). After analyzing the BCI systems at home and abroad in recent years, this paper studies the BCI system based on event related potentials (ERP). A novel input system based on event-related potentials is designed by using a non-invasive EEG based BCI research method, which can input codes directly through EEG signals without hand. The contents of this paper include the design of the experimental paradigm of input system, the design and implementation of the whole system, the analysis and processing of EEG signal, etc. In the design of experimental paradigm, the form and color of input characters are compared according to the potential characteristics of P300 in ERP, and on this basis, different experimental parameters are used to carry out many experiments. In this paper, according to the experimental paradigm of the implementation of the system, mainly using MFC in VC programming. The system integrates the generation of stimulation interface, the collection, storage and analysis of EEG signals, and can output the target characters of the subjects to the display interface in real time. Based on the analysis of the P300 data in the third BCI contest, the processing scheme of the self-collected EEG signal is determined in this paper. For the competition data, the filter is first used to filter, the downsampling and time window are used to reduce the dimension, and then Fisher linear discriminant, regularization Fisher, kernel Fisher, support vector machine and Bayesian linear discriminant are used to classify the competition data. Results the surface regularization Fisher achieved the highest accuracy with less stacking times in a short time. The effect of Fisher linear discriminant was the second. Therefore, after using superposition average and independent component analysis (ICA) to remove interference from self-collected data, Fisher linear discriminant and regularized Fisher are used to classify the data. Finally, the optimal input mode, experimental parameters and EEG signal processing scheme are determined.
【学位授予单位】：华侨大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP334.7;TN911.7

【相似文献】