模拟阅读脑—机接口信号处理与分类

发布时间：2018-01-14 12:14

本文关键词：模拟阅读脑—机接口信号处理与分类　出处：《中南民族大学》2012年硕士论文　论文类型：学位论文

【摘要】：脑-机接口技术是一种不依赖于外周神经和肌肉组织等大脑的正常的输出通道的通信系统。它为那些神经或肌肉组织受损而不能使用正常通信手段的患者，提供了一种直接从大脑获取与外界进行通信交流的信息，从而实现对计算机等外部设备的控制，提高了患者的生活质量。此外，脑-机接口技术在人机自动控制、军事领域等领域也有潜在的研究价值。脑-机接口系统中的核心问题是一种翻译算法，如何把来自使用者的电生理信号（比如脑电信号）转换成输出控制信号，从而控制外部设备。本论文通过查阅文献并总结国际上脑-机接口研究的现状，以及脑电信号处理的各种方法，从以下几点进行研究。 ⑴脑电信号采集的实验模式及预处理。文中采用的实验模式是由中南民族大学生物医学工程实验室设计的基于模拟阅读诱发脑电采集模式。该模式的受试者像平常阅读那样在不带突变的物理刺激的靶标符号中获取视觉信息。又因为脑电信号很微弱并且极容易受到各种干扰很热噪声的影响，所以文中利用模拟阅读的实验模式很好地降低了眼动伪迹干扰的污染，利用低通滤波、下采样等处理进一步滤除污染脑电信号的干扰和噪声。实验表明效果理想。 ⑵脑电信号特征量的提取。有效的提取大脑思维活动中意识信息是脑-机接口技术研究的关键技术之一，是正确识别不同意识模式的基础。文中在参考众多脑电信号特征提取方法研究的基础上采用共同空间模式（Common Spatial Pattern，CSP）方法对预处理后的脑电信号进行特征提取，该方法的目标是设计空间滤波器，目的是在高维空间中给出两个分类，找到一个最大化的第一类方差，同时最小化第二类方差。实验表明利用此方法提取的脑电信号特征对后续的分类识别有明显的效果。 ⑶脑电信号分类器的设计。分类器的设计是脑-机接口系统中另一个重要环节，分类器的性能直接影响着脑-机接口系统的性能。文中利用支持向量机（Support Vector Machine，SVM）方法对脑电信号进行分类识别，该算法是以统计学习理论为基础，经验最小化原则为前提寻求既能准确将两类样本数据分开，又能使类间的分类间隔最大化的最优分类面。实验分类结果表明该方法脑电信号的分类识别有很好的效果。最后文中对基于模拟阅读脑电信号在通过以上步骤的处理后进行分类识别，结果表明分类识别的效果理想。文中对5名受试者的进行实验并对所采集的脑电信号进行分类识别，在实验中对每名受试者进行随机的15次重复试验，，得出每名受试者每一次的分类准确率，并统计它们的平均准确率。试验结果显示最高的分类率达到97%，平均准确率最高也能达到90%以上，因此通过实验结果可得出，文中基于CSP和SVM相结合的方法适合于基于模拟阅读的脑-机接口中脑电信号的研究。
[Abstract]:The brain - computer interface technology is a communication system which does not rely on the normal output channel of the brain such as peripheral nerve and muscle tissue . The core problem in the brain - computer interface system is a translation algorithm , how to convert electrophysiological signals from users ( such as EEG signals ) into output control signals , so as to control external devices . ( 1 ) Experimental mode and pretreatment of EEG signal acquisition . The experimental model used in this paper is designed by the Biomedical Engineering Laboratory of the National University of China and the South , which is based on the simulated reading evoked brain electrical acquisition mode . The subjects of the model can obtain the visual information in the target symbols without abrupt physical stimulation as usual . Because the EEG signals are weak and are very easy to be influenced by various disturbances , the experimental mode of the simulated reading is very good for reducing the pollution of eye movement artifacts , and the interference and noise of the polluted brain electrical signals are further filtered by low - pass filtering and down - sampling , and the experiment shows that the effect is ideal . ( 2 ) extracting the characteristic quantity of the brain electrical signal . It is one of the key techniques of brain - computer interface technology research to extract the consciousness information in brain - computer interface technology . Based on the research of the feature extraction methods of EEG signals , a spatial filter is designed . The goal of this method is to design a spatial filter . The aim is to design a spatial filter . The aim is to design a spatial filter . The aim is to design a space filter . The aim is to design a space filter . The aim is to find a maximum first type variance and to minimize the second type variance . The experiment shows that the EEG signal extracted by this method has obvious effect on the subsequent classification . ( 3 ) Design of EEG signal classifier . The design of the classifier is another important part of the brain - computer interface system . The performance of the classifier directly affects the performance of the brain - computer interface system . This algorithm is based on the statistical learning theory and the experience minimization principle . The results show that the highest classification rate is 97 % and the average accuracy rate can reach more than 90 % . The results show that the method based on CSP and SVM is suitable for the study of EEG signal in brain - machine interface based on simulated reading .

【学位授予单位】：中南民族大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP334.7;R318.0

【参考文献】