【作者】 许涛；

【导师】 朱林剑；

【作者基本信息】 大连理工大学 ， 生物医学工程， 2006， 硕士

【摘要】 基于脑电信号(Electroencephalogram，EEG)的假肢系统是用人的意愿实现人脑与外部电子设备通讯和控制的系统，是生物医学工程学中的重要分支，其研究涉及神经生理学、康复工程学和信号处理技术等各学科。其中脑电信号的特征信息提取和分类识别在脑电假肢系统中扮演着十分重要的角色。 目前，EEG除了用于医学研究，主要集中在为残疾人提供与外界交流的途径以恢复某些肢体功能。在EEG控制假肢的研究中发现：作为EEG控制假肢系统的信号处理部分左右了整个系统的性能，是系统性能好坏的主要因素。然而EEG本身及其复杂，在识别上有很大的困难，不能依据采集到的EEG直接控制假肢的动作，选择合适的特征提取方法和分类算法显得尤为重要。在学习并总结了国内外此领域内的大量文献资料的基础上，以探索脑电假手的可行性和降低脑电假手的误动率为目的，结合EEG节律特点，采用小波包分解和BP神经网络相结合的脑电假手控制方案。 首先，对EEG进行分析，不同状态的EEG在节律上表现出不同模式。在此基础上，总结了前人应用EEG作为控制信号的实践经验，在实验的基础上分析了应用EEG控制假手的可行性，并选择利用意识任务EEG控制假手动作。 其次，研究了小波变换的基本理论，利用小波变换的特点，对EEG进行滤波降噪处理。提出根据EEG节律提取相应频带能量特征的方法，利用小波包变换的多分辨率分析对EEG进行分解，选择与EEG节律对应的频率子信号，重构并计算出能量特征，作为表征该状态EEG的特征信息。 第三，深入分析了BP神经网络的结构特点并对其进行改进，通过对实验数据的分析，对BP神经网络各参数进行合理设置，并利用其自适应、自学习的特点对不同状态EEG的特征信息进行分类识别，输出控制指令。 最后，利用电路设计的相关知识对外部控制电路进行设计，进行简单的假手动作实验。以图形化语言LabVIEW7为开发平台实现PC机和外部控制电路的串口通信，并进行了脑电信号显示界面的设计，实现了信号显示和处理功能。更多还原

【Abstract】 The artificial hands based on EEG is the system for controlling or communicating with other electronic devices by human intentions. It is the important branch of the Biomedical Engineering. The research deals with the neurophysiology, rehabilitation engineering, signal process technology and so on. The electroencephalograph(EEG) feature extraction and EEG classification play important roles.So far, the studies in EEG have been concentrated on providing an efficient way to communicate with the outer-world and recall some limb functions for individuals with severe motor defects. In the research of artificial hands controlled by EEG, finding that the EEG identification correctly plays the key role in the artificial hands system. EEG is very complicated biologic signal. The EEG collected directly isn’t of enough information and can’t adapt to the development of artificial hand. So it is very important to choose correct methods of EEG feature extraction and classification. For improving the precision of motion, a kind of control model is devised based on wavelet packet transformation and BP algorithm after summing up and studying other works in this field in this paper.First, study EEG deeply and other works in this field. By these preparative, the author analyses the feasibility of control artificial with EEG and devises a control model controlling artificial hand with mental tasks EEG.Second, research on wavelet transformation theory and filter noises from EEG using it. Furthermore, research on wavelet packet transformation. Multiresolution analysis theory is used to extract different rhythm power features which represent the corresponding EEG based on EEG rhythm.Third, the structure of BP neural network(NN) is studied and also improved. Parameters of NN are set correctly. The different EEG is classified by devised NN which is self-adjusting and self-learning. Control command is outputted after classification by NN.Finally, a simple artificial hands motion experiment is done using the knowledge of digital and simulation circuit. Finish the serial communication from PCs to outer control circuit on the development environment LabVIEW and design a brain-computer interface of EEG displaying which can do the simple EEG processing.更多还原