基于运动想象EEG信号的特征提取研究

发布时间：2018-04-28 01:31

本文选题：脑-机接口 + 脑电图　；参考：《南京邮电大学》2017年硕士论文

【摘要】：脑-机接口是一种全新的人工智能控制系统,使用户与电子设备进行交互的信息通路发生了改变,创造了一种全新的交互方式。脑-机接口系统在辅助康复领域有巨大的应用空间。当人类在进行思维活动时,大脑会产生各种信号,这些信号可以通过脑电、脑磁、fMRI等多种途径记录下来。与其他信号相比,脑电信号具有制造成本低、时间分辨率高、小巧便携等优势。论文基于脑电信号,对基于运动想象任务的脑-机接口系统进行了研究,重点研究了各种特征提取技术。论文首先对大脑的结构和脑电图信号的产生机制、类型以及脑-机接口系统的研究方法进行了介绍,重点介绍了常用特征提取、分类方法,并分析了不同算法的优缺点。其次,设计了一个基于运动想象任务的脑电图信号采集实验。论文详细介绍了实验方案、实验设备、实验过程以及实验的注意事项。然后,论文运用排列组合熵特征,研究了排列组合熵作为基于运动想象的脑-机接口系统的特征的可行性,并对其进行改进,提出了一种基于EMD分解重构的排列组合熵算法。使用该算法提取排列组合熵特征,并使用线性判别分类器进行分类,分类正确率相比原始的排列组合熵算法有了提高,说明了该算法的有效性。论文还提出了一种基于改进的频带能量的特征提取算法。首先对频带能量算法进行研究分析,分析了其优缺点,提出了一种对当前时刻以前的频带能量值进行累积的特征提取算法,该算法提高了分类正确率和最大互信息这两个评价指标,将结果与使用该数据的其他研究人员的研究成果进行比较,该算法有着明显的优势。
[Abstract]:Brain-computer interface (BCI) is a new artificial intelligence control system, which changes the information path between users and electronic devices and creates a new way of interaction. The brain-computer interface system has great application space in the field of auxiliary rehabilitation. When people are thinking, the brain produces various signals, which can be recorded by EEG, MRI and other channels. Compared with other signals, EEG signals have the advantages of low cost, high time resolution and small portability. Based on EEG, the brain-computer interface system based on motion imagination task is studied in this paper, with emphasis on various feature extraction techniques. Firstly, the structure of the brain and the mechanism of EEG signal generation, the types and the research methods of brain-computer interface system are introduced. The common feature extraction and classification methods are introduced, and the advantages and disadvantages of different algorithms are analyzed. Secondly, an EEG signal acquisition experiment based on motion imagination task is designed. This paper introduces the experimental scheme, experimental equipment, experimental process and points for attention in detail. Then, the feasibility of permutation and combination entropy as the feature of brain-computer interface system based on motion imagination is studied and improved, and a permutation and combination entropy algorithm based on EMD decomposition and reconstruction is proposed. The algorithm is used to extract the feature of permutation combination entropy and to use linear discriminant classifier to classify. The classification accuracy is improved compared with the original permutation combination entropy algorithm, which shows the validity of the algorithm. A feature extraction algorithm based on improved band energy is also proposed. Firstly, the band energy algorithm is studied, and its advantages and disadvantages are analyzed, and a feature extraction algorithm is proposed, which accumulates the frequency band energy value before the current moment. The algorithm improves the classification accuracy and the maximum mutual information, and compares the results with the research results of other researchers who use the data. The algorithm has obvious advantages.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN911.7;R318

【参考文献】