基于LM算法的脑电信号分类研究  

Research on the classification of EEG signals based on LM algorithm

作　　者：赵东东 宋洪军 许玉虎 崔东云 王帅 丁筱玲[1] Zhao Dongdong;Song Hongjun;Xu Yuhu;Cui Dongyun;Wang Shuai;Ding Xiaoling(College of Mechanical and Electronic Enginerring,Shandong Agricultural University,Taian 271018,China;SDIC Fund Management Company Ltd.,Beijing 100000,China)

机构地区：[1]山东农业大学机械与电子工程学院,山东泰安271018 [2]国投创新投资管理有限公司,北京100000

出　　处：《电子技术应用》2018年第12期20-24,共5页Application of Electronic Technique

摘　　要：为实现运动想象脑电信号的精准分类,提出以Levenberg-Marquardt算法(LM)替代BP神经网络构造分类器来提高分类识别率。实验以2008年BCI竞赛信号采集模式为标准,使用Emotive Epoc+采集四类运动想象脑电信号,对采集的信号进行滤波去燥后,利用主成分分析提取特征值;然后分别用LM算法和BP神经网路进行分类识别做对比;最后基于MATLAB GUI设计串口通信界面与Arduino智能车链接验证算法的可行性。结果证明:该方法训练平均误差为5.630 6×10^(-7),分类准确率为86%,BP算法相对应为0.001 4、56%。相对比可知LM算法分类效果良好,验证过程中,智能车运行与算法识别方向一致,运行良好。此方法切实可行,为后期进一步开发脑机接口奠定了基础。In order to realize the accurate classification of EEG signals based on motion imagination,the Levenberg Marquardt(LM)algorithm is proposed to replace the BP neural network classifier to improve the classification recognition rate.Based on the BCI2008competition laboratory paradigm,we used Emotive Epoc+to collect four kinds of motor imagery EEG signals.After filtering the signal to dryness,the main component analysis is used to extract the characteristic values of the collected signals,and then the LM algorithm and the BP neural network are used for classification and recognition respectively.Finally,the serial communication interface is designed based on the MATLAB GUI,and the feasibility of the algorithm is verified with the Arduino intelligent car link.The results show that the average training error is5.6306×10^-7,the classification accuracy is86%,and the BP algorithm is0.0014and56%respectively.Compared with the LM algorithm,the classification effect is good.During the verification process,the intelligent car operation is consistent with the algorithm identification direction,and runs well.This method is practical and feasible,which lays the foundation for further developing brain computer interface.

关 键 词：脑电信号 运动想象 BP神经网络 LM算法 MATLAB GUI 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]