基于Adaboost的BCI系统脑电信号分类

发布时间：2018-01-25 22:36

本文关键词： BCI系统脑电信号 CSP滤波 PCA AdaboostNN　出处：《天津大学》2012年硕士论文　论文类型：学位论文

【摘要】：脑机接口是建立在人脑和外部设备之间的直接通讯通路。脑机接口技术结合了生物医学、计算机信息处理技术、神经科学以及微电子等多个领域的最新成果,在近10年的时间里得到的广泛的重视研究和发展。通过对脑电信号的研究,脑机接口BCI(Brain-Computer Systerm)系统已经可以被用来解决很多实际问题。尽管如此,因为脑电信号的不稳定性和个体差异性,找到一种高效的、具有普遍意义的信号处理和识别方式是解决问题的关键。本文的主要研究对象是BCI系统中脑电信号的处理和识别方法。通常一个模式识别过程可以分为数据预处理、特征提取、特征选择和降维、以及特征分类几个阶段。在脑电信号的预处理阶段使用FIR数字滤波器和CSP空域滤波器的方法对脑电信号进行滤波处理,用主成分分析PCA和偏最小二乘PLS的方法对脑电信号的特征向量进行降维,用两种不同的Adaboost分类方法对提取到的脑电信号特征向量进行分类,同时用线性判别分写LDA和支持向量机SVM的BCI系统经典分类方法进行分类识别率的对比。通过对在模式识别每一个阶段对脑电信号处理的不同方法的研究,在实验结果数据的基础上给出一种高效的处理脑电信号的方法,即公共空间模式滤波与以最近邻法为若分类其的AdaboostNN分类器结合的处理方式。
[Abstract]:Brain-computer interface (BCI) is a direct communication pathway between human brain and external devices. BCI technology combines the latest achievements in many fields such as biomedicine, computer information processing, neuroscience, microelectronics and so on. In the past 10 years, extensive attention has been paid to research and development through the study of EEG signals. Brain-Computer Interface (BCI(Brain-Computer Systerm) systems have been used to solve many practical problems. Because of the instability and individual difference of EEG signal, we find a kind of high efficiency. The main research object of this paper is the method of EEG signal processing and recognition in BCI system. Usually a pattern recognition process can be divided into data. Pretreatment. Feature extraction, feature selection and dimensionality reduction, and feature classification. The FIR digital filter and CSP spatial filter are used to filter the EEG signal in the preprocessing stage. The method of principal component analysis (PCA) and partial least squares (PLS) is used to reduce the dimension of the Eigenvectors of EEG. Two different Adaboost classification methods are used to classify the extracted feature vectors of EEG signals. At the same time, the classical classification methods of BCI system based on linear discriminant classification (LDA) and support vector machine (SVM) are compared. Through the research on different methods of EEG processing in each stage of pattern recognition, an efficient method of EEG signal processing is presented on the basis of experimental data. That is the combination of common space mode filtering and AdaboostNN classifier whose nearest neighbor method is the nearest neighbor method.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP334.7

【参考文献】