用于生物特征识别的多范式诱发脑电个体差异性研究
发布时间：2019-03-11 08:56
【摘要】：近年来,生物特征识别技术受到全世界各国的普遍关注,它在维护国家安全、个人信息安全、航空安全以及军事、医疗等方面均发挥着重要的作用,成为信息化时代的前沿热点课题。为了满足特殊应用场景的需求和弥补现有技术的不足,研究者们始终致力于开发新的生物特征,脑电是其中较为新颖的尝试之一。基于脑电的生物特征识别技术研究在国内外尚属起步阶段,在脑电诱发范式的设计、脑电特征提取以及模式识别算法等方面仍存在较大的探索空间。 本研究从脑电的个体差异性出发,首先针对现有研究中脑电诱发范式相对单一的问题,设计并完成了涵盖多种脑电诱发范式的两类事件任务实验:第一类任务包括静息、视觉认知、计算任务和运动想象四种范式,样本量20人;第二类任务是视觉诱发P3范式,样本量8人。为了有效地提取脑电中的个体差异性信息,针对各范式诱发脑电的信号特点,研究中分别尝试了AR模型、时域能量谱、频域能量谱、相位锁定值以及相干平均等多种时域和频域的特征提取算法,并利用支持向量机对上述各种范式的脑电特征进行分类识别,从而得到基于全样本的统计分类正确率。本文研究结果表明,诱发脑电的个体差异性明显高于静息脑电,而且越是复杂任务的、被试参与度高的、与思维活动密切相关的范式,其诱发的脑电个体差异性越明显,分类正确率最高可达98%以上,从而验证了脑电可用于生物特征识别的可行性。 在此基础上,为了进一步提高系统性能和识别效率,本研究在特征筛选和导联优化方面做了初步探索,并通过遗传算法、Fisher判别率和递归特征筛选三种方法对分类器进行了优化,与优化前相比不仅识别率有所提高,而且导联数得到明显简化,从而为今后脑电的个体差异性分析及其面向生物特征识别的实用性设计提供了新的思路。
[Abstract]:In recent years, biometric identification technology has been widely concerned all over the world. It plays an important role in maintaining national security, personal information security, aviation security, military affairs, medical treatment and so on. It has become a hot topic in the information age. In order to meet the needs of special application scenarios and make up for the shortcomings of existing technologies, researchers have always been devoted to the development of new biological characteristics, among which EEG is one of the more novel attempts. The research of EEG-based biometric recognition technology is still in its infancy at home and abroad. There is still a large exploration space in the design of EEG-induced paradigm, EEG feature extraction and pattern recognition algorithm. Based on the individual differences of EEG, this study designs and completes two kinds of event task experiments covering multiple EEG evoked paradigms: the first type of tasks includes resting tasks, aiming at the problem that EEG evoked paradigm is relatively single in the existing studies, and two kinds of event task experiments covering various EEG evoked paradigms are designed and completed. Visual cognition, computational tasks and motor imagination four paradigms, sample size of 20; The second type of task was the visual evoked P3 paradigm, with a sample size of 8. In order to extract the individual difference information effectively, the AR model, the time domain energy spectrum and the frequency domain energy spectrum are tried in the study, according to the signal characteristics of the evoked EEG in each normal form. A variety of time-domain and frequency-domain feature extraction algorithms such as phase-locked values and coherent averages are used to classify and recognize EEG features from the above-mentioned paradigms using support vector machines to obtain the statistical classification accuracy rate based on full samples. The results of this study show that the individual difference of evoked EEG is significantly higher than that of resting EEG, and the more complex tasks, the higher the participation of subjects and the paradigm closely related to thinking activity, and the more obvious the individual difference of evoked EEG is, the more complex the task is, the higher the participants' participation is, and the more closely related to thinking activities. The correct rate of classification is up to 98%, which proves the feasibility of EEG in biometric recognition. On this basis, in order to further improve the system performance and recognition efficiency, this study has made a preliminary exploration in feature screening and lead optimization, and through genetic algorithm, Three methods of Fisher discrimination rate and recursive feature selection are used to optimize the classifier. Compared with the pre-optimization, the recognition rate is improved and the lead number is obviously simplified. It provides a new idea for the analysis of individual difference of EEG and the practical design for biometric recognition.
【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】：R318.0
本文编号：2438149
[Abstract]:In recent years, biometric identification technology has been widely concerned all over the world. It plays an important role in maintaining national security, personal information security, aviation security, military affairs, medical treatment and so on. It has become a hot topic in the information age. In order to meet the needs of special application scenarios and make up for the shortcomings of existing technologies, researchers have always been devoted to the development of new biological characteristics, among which EEG is one of the more novel attempts. The research of EEG-based biometric recognition technology is still in its infancy at home and abroad. There is still a large exploration space in the design of EEG-induced paradigm, EEG feature extraction and pattern recognition algorithm. Based on the individual differences of EEG, this study designs and completes two kinds of event task experiments covering multiple EEG evoked paradigms: the first type of tasks includes resting tasks, aiming at the problem that EEG evoked paradigm is relatively single in the existing studies, and two kinds of event task experiments covering various EEG evoked paradigms are designed and completed. Visual cognition, computational tasks and motor imagination four paradigms, sample size of 20; The second type of task was the visual evoked P3 paradigm, with a sample size of 8. In order to extract the individual difference information effectively, the AR model, the time domain energy spectrum and the frequency domain energy spectrum are tried in the study, according to the signal characteristics of the evoked EEG in each normal form. A variety of time-domain and frequency-domain feature extraction algorithms such as phase-locked values and coherent averages are used to classify and recognize EEG features from the above-mentioned paradigms using support vector machines to obtain the statistical classification accuracy rate based on full samples. The results of this study show that the individual difference of evoked EEG is significantly higher than that of resting EEG, and the more complex tasks, the higher the participation of subjects and the paradigm closely related to thinking activity, and the more obvious the individual difference of evoked EEG is, the more complex the task is, the higher the participants' participation is, and the more closely related to thinking activities. The correct rate of classification is up to 98%, which proves the feasibility of EEG in biometric recognition. On this basis, in order to further improve the system performance and recognition efficiency, this study has made a preliminary exploration in feature screening and lead optimization, and through genetic algorithm, Three methods of Fisher discrimination rate and recursive feature selection are used to optimize the classifier. Compared with the pre-optimization, the recognition rate is improved and the lead number is obviously simplified. It provides a new idea for the analysis of individual difference of EEG and the practical design for biometric recognition.
【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】：R318.0
【参考文献】
相关期刊论文 前5条
1 杨艳;;生物特征识别技术概述[J];大众科技;2006年08期
2 田启川;张润生;;生物特征识别综述[J];计算机应用研究;2009年12期
3 卢官明;李海波;刘莉;;生物特征识别综述[J];南京邮电大学学报(自然科学版);2007年01期
4 孔薇,杨杰,周越;基于独立成分分析的强背景噪声去噪方法[J];上海交通大学学报;2004年12期
5 高上凯;;浅谈脑—机接口的发展现状与挑战[J];中国生物医学工程学报;2007年06期
相关博士学位论文 前2条
1 程龙龙;BCICFES重建运动神经系统的信号处理与控制关键技术研究[D];天津大学;2010年
2 綦宏志;脑—机接口中上肢想象动作侧向性识别的关键技术研究[D];天津大学;2008年
相关硕士学位论文 前4条
1 张广举;基于膝关节角度的助行功能性电刺激模糊控制研究[D];天津大学;2010年
2 郭晓静;基于独立分量分析的思维脑电和诱发脑电的特征提取[D];安徽大学;2003年
3 邹清;基于AR模型的脑电信号特征提取与识别[D];中南大学;2008年
4 刘俊俏;基于运动想象的脑信号处理[D];华南理工大学;2010年
，本文编号：2438149
本文链接：https://www.wllwen.com/yixuelunwen/swyx/2438149.html
