脑机接口中的运动想象脑电信号分析与处理方法研究

发布时间：2018-01-26 11:16

本文关键词： 脑电信号自适应小波阈值经验模态分解模糊熵人工蜂群支持向量机　出处：《杭州电子科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：脑-机接口(Brain-Computer Interface,BCI)是一种基于脑电信号(Electroencephalogram,EEG)的人机交互方式,它能够不需要依靠周边的肌肉和神经组织,只依靠脑部信号便可以达到自由行动及与外界沟通等目的。本文以基于运动想象的脑电信号为研究对象,对EEG信号预处理、特征提取和模式分类等处理过程进行研究,本文主要完成的研究工作如下:(1)预处理:为了减少信号中的噪声和其它的干扰成分,本文利用自适应的小波阈值去噪算法对采集的脑电信号进行处理。实验采用BCI国际竞赛的数据集,结果表明该方法在很大程度上减少了噪声干扰。(2)特征提取:由于脑电信号的非线性特点,本文提出了将改进的经验模态分解方法与模糊熵算法相结合的特征提取方法。原始的脑电信号经过分解得到若干个固有模态分量(IMF),然后利用互信息筛选出有用的IMF分量并进行重构,最后利用模糊熵对其进行特征提取,并采用2008年的BCI竞赛数据验证算法的特征提取效果。(3)模式分类:当采用支持向量机(SVM)进行模式分类时,核参数g和惩罚因子C的选择关系着分类器的性能,所以本文利用人工蜂群算法(ABC)优化SVM的模型参数,以增强分类器的分类性能,并对提取的脑电信号特征进行模式分类。(4)实验结果与分析:实验采用2005年和2008年BCI国际竞赛的数据集,实验一对比了ABC算法与传统优化算法的寻优性能,证明了ABC算法具有良好的寻优性能。实验二的结果表明ABC算法优化后增强了SVM的分类性能,有效地提高了脑电信号的分类正确率。实验三利用改进的EMD和模糊熵算法进行特征提取,然后结合人工蜂群优化的支持向量机对提取的特征进行分类,由实验结果可知,改进EMD和模糊熵算法相结合的方法是有效的,经ABC优化的SVM得到的识别率比传统的SVM要高。
[Abstract]:Brain-Computer Interface (BCI) is a kind of electroencephalogram based on electroencephalogram (EEG). EGG is a human-computer interaction that allows it to rely on peripheral muscles and neural tissue without needing to rely on it. We can only rely on the brain signal to achieve the goal of free movement and communication with the outside world. In this paper, we preprocess the EEG signal with the EEG signal based on motion imagination as the research object. The main work of this paper is as follows: preprocessing: in order to reduce the noise and other interference components in the signal. In this paper, we use the adaptive wavelet threshold denoising algorithm to process the collected EEG signals. The experiment adopts the data set of BCI international competition. The results show that this method greatly reduces the noise interference. In this paper, an improved empirical mode decomposition (EMD) method combined with fuzzy entropy algorithm is proposed. The original EEG signal is decomposed to obtain a number of intrinsic modal components (IMF). Then the useful IMF components are selected by mutual information and reconstructed, and the feature extraction is carried out by fuzzy entropy. And the feature extraction effect of the BCI contest data validation algorithm of 2008 is used to classify the pattern: when using support vector machine (SVM) to classify the pattern. The selection of kernel parameter g and penalty factor C is related to the performance of classifier, so the artificial bee colony algorithm is used to optimize the model parameters of SVM in order to enhance the classification performance of the classifier. The experimental results and analysis were as follows: the data sets of BCI international competition on 2005 and 2008 were used in the experiment. Experiment 1 compares the optimization performance of ABC algorithm with that of traditional optimization algorithm, and proves that ABC algorithm has good optimization performance. Experiment 2 shows that ABC algorithm improves the classification performance of SVM after optimization. Experiment 3 uses improved EMD and fuzzy entropy algorithm to extract features, and then combines the support vector machine of artificial bee colony optimization to classify the extracted features. The experimental results show that the combination of improved EMD and fuzzy entropy algorithm is effective, and the recognition rate of SVM optimized by ABC is higher than that of traditional SVM.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R318;TN911.7

【参考文献】