具有在线自学习能力的脑电信号分类方法  

EEG Signal Classification Method with Online Self-learning Ability

作　　者：李红宇[1] 刘庆江[1] 常晓娟[1] 赵薇[1] Li Hongyu;Liu Qingjiang;Chang Xiaojuan;Zhao Wei(Harbin Normal University,Harbin Heilongjiang 150025,China)

机构地区：[1]哈尔滨师范大学

出　　处：《信息与电脑》2019年第23期99-100,103,共3页Information & Computer

基　　金：黑龙江省教育厅科学技术研究项目“一种自适应脑电信号分类方法研究”（项目编号：12541255）

摘　　要：脑电信号具有时变、个体差异的特点,容易受到身体状态、情绪、位置等因素的影响,传统的BP网络分类器难以适应动态监测的要求。基于此,笔者提出了一种基于BP网络的BP AdaBoost基本网络分类器。首先,该分类器是在传统Ada在线自学习能力的脑电信号分类方法boost集成学习框架下由弱分类器形成的,其通过引入遗忘因子,改变初始样本容量来改进AdaBoost算法;其次,初始权值增强了其时间相关性,得到BP-AdaBoost分类器,并进一步借鉴半监督的思想,增加了基于K-近邻规则的自评价反馈环节,从而提高了捕获效果;最后,基于国际BCI竞赛数据集,利用Hilbert Huang变换提取脑电图特征。仿真结果表明,笔者提出的分类方法对时间和个体具有较好的适应性和鲁棒性,与传统的BP神经网络相比,分类精度约提高了23.42%。EEG signals have the characteristics of time-varying and individual differences,and are easily affected by factors such as physical state,emotion,and location.Traditional BP network classifiers are difficult to adapt to the requirements of dynamic monitoring.Based on this,the authors propose a BP AdaBoost basic network classifier based on BP network.First,the classifier is formed by a weak classifier under the boost integrated learning framework of the traditional ADA online self-learning capability EEG signal classification method.,it improves the AdaBoost algorithm by introducing forgetting factors and changing the initial sample capacity;second,the initial weight Value enhances its time correlation,obtains the BP-AdaBoost classifier,and further draws on the semi-supervised idea,adds a self-evaluation feedback link based on the K-nearest neighbor rule,thereby improving the capture effect;finally,based on the international BCI competition dataset Hilbert Huang transform was used to extract EEG features.Simulation results show that the classification method proposed by the author has good adaptability and robustness to time and individuals.Compared with the traditional BP neural network,the classification accuracy is improved by about 23.42%.

关 键 词：自主学习 脑电图 识别 学习演算法 

分 类 号：TN911.7[电子电信—通信与信息系统]