P300脑机接口系统的范式设计及其算法研究

发布时间：2019-03-17 21:01

【摘要】：脑-机接口(Brain-Computer Interface, BCI)技术不依赖于大脑的外周神经系统与肌肉组织为大脑与外界开辟了一条新的特殊通道。它可以不需要语言或肢体动作而直接通过脑电来完成对外部环境及装置的控制。当前的脑-机接口研究正处于发展阶段,如何快速准确地识别脑电模式,是脑-机接口领域中的一个热点问题。本文展开了基于P300的脑机接口系统的研究。P300是一种小概率诱发电位,常用于构建脑-机接口系统的脑电信号。P300响应是在目标刺激300ms左右出现的一个正向波。基于P300的脑-机接口系统,受试者不需要进行特别的训练就能达到较好的效果。本文工作是对P300实验范式和分类识别的研究。主要做了以下几项工作： (1)选择贝叶斯线性判别分析(BLDA),通过处理离线阶段和在线阶段的实验数据,研究随机激励和非随机激励产生的P300信号的特点,得出P300随机激励诱发产生的P300比非随机激励更明显、易于识别。 (2)分别用贝叶斯线性判别分析(BLDA)、线性判别分析(LDA4)与支持向量机(SVM)作为分类识别算法,对离线阶段随机激励产生的P300信号分类结果进行比较。基于BLDA算法的分类精度最优；LDA4算法下也得到比较好的分类精度,相对BLDA的效果次之；SVM下的分类精度相对比较差。且SVM识别所花费比较长的时间,BLDA所花费时间最短,相对于SVM快很多。 (3)在脑电信号的消噪上,引进了独立分量分析(ICA)方法,通过Infomax ICA和FastICA两种算法消除脑电信号中的眼电,以期得到噪声较少的脑电信号。利用贝叶斯线性判别分析(BLDA)方法,来判别Infomax ICA和FastICA去噪效果,并且和未去噪的结果进行比较,来验证去噪的有效性。
[Abstract]:Brain-computer interface (Brain-Computer Interface, BCI) technology does not depend on the peripheral nervous system and muscle tissue of the brain to open up a new special channel for the brain and the outside world. It can control the external environment and device directly through EEG without the need of language or limb movement. At present, the research of brain-computer interface is in the developing stage. How to recognize EEG pattern quickly and accurately is a hot issue in the field of brain-computer interface. In this paper, a P300-based brain-computer interface system is studied. P300 is a small probability evoked potential, which is often used to construct EEG signals of brain-computer interface system. P300 response is a positive wave around the target stimulation 300ms. P300-based brain-computer interface system, the subjects do not need special training to achieve better results. The work of this paper is to study the P300 experimental paradigm and classification recognition. The main work is as follows: (1) the Bayesian linear discriminant analysis (BLDA),) is selected to study the characteristics of P300 signals generated by random excitation and non-random excitation by processing the experimental data of off-line and on-line phases. It is concluded that P300 induced by random excitation is more obvious than non-random excitation and is easy to identify. (2) using Bayesian linear discriminant analysis (LDA4) (BLDA), linear discriminant analysis (LDA4) and support vector machine (SVM) as classification recognition algorithms, the classification results of P300 signals generated by off-line random excitation are compared. The classification accuracy based on BLDA algorithm is the best; the classification accuracy under LDA4 algorithm is better than that of BLDA; the classification accuracy under SVM is relatively poor. And SVM takes a long time to identify, and BLDA takes the shortest time, much faster than SVM. (3) Independent component analysis (ICA) is introduced to eliminate the noise of EEG signals. The two algorithms, Infomax ICA and FastICA, are used to eliminate the eye electricity in EEG signals, in order to get the EEG signals with less noise. The Bayesian linear discriminant analysis (BLDA) method is used to judge the denoising effect of Infomax ICA and FastICA, and the results are compared with those of non-denoising to verify the effectiveness of de-noising.
【学位授予单位】：华东理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP334.7

【参考文献】