Please wait a minute...
浙江大学学报(工学版)
J4  2013, Vol. 47 Issue (8): 1431-1436    DOI: 10.3785/j.issn.1008-973X.2013.08.016
电气工程     
脑机接口中基于BISVM的EEG分类
杨帮华, 何美燕, 刘丽, 陆文宇
上海大学 机电工程与自动化学院自动化系; 上海市电站自动化技术重点实验室,上海 200072
EEG classification based on batch incremental SVM in
brain computer interfaces
YANG Bang-hua, HE Mei-yan, LIU Li, LU Wen-yu
Department of Automation, School of Mechatronics Engineering and Automation; Shanghai Key Laboratory of Power Station Automation Technology; Shanghai University, Shanghai 200072
 全文: PDF 
摘要：

针对脑电信号(EEG)分类问题,提出基于批处理增量式支持向量机(BISVM)的分类方法.将所有数据通过批处理进行分组,采用第1组数据在SVM中建立初始分类器模型,将剩余组内数据顺序作为新增样本,对满足卡罗需-库恩-塔克(KKT)条件的样本进行增量学习和减量去学习,不断判断KKT条件并更新参数,丢弃错误样本,对初始分类器模型进行更新.对2008年脑机接口竞赛数据及本实验室采集数据,用小波包分解(WPD)结合共空间模式(CSP)进行特征提取,SVM、ISVM及BISVM分类.结果表明，BISVM的平均分类准确率相对SVM及ISVM分别提高了3.3%及0.3%,BISVM平均训练时间相对ISVM从1.076 s减少到0.793 s.BISVM为改善计算机对大脑的适应性,实现快速实时在线的脑机接口系统奠定基础.
关键词： 脑机接口;  批处理增量式支持向量机;  脑电;  分类    
Abstract:

Aiming at the Electroencephalogram (EEG) classification, a classification method based on batch incremental support vector machine (BISVM) was proposed in brain computer interfaces (BCIs). All training data were grouped by batch processing and the initial SVM classifier model was set up using the first group. The remaining groups of data were added into the first group orderly as new samples. An incremental learning and decremental learning process was adopted to samples which meet the KKT condition. The initial SVM classifier model was updated by continuously estimating KKT condition, updating the parameters, discarding the error samples. Based on the 2008 BCI Competition Dataset and our experimental EEG data, features were extracted by wavelet packet decomposition (WPD) and common spatial patterns (CSP). The SVM, ISVM and BISVM were used to classify these features. Experimental results show that the average classification accuracy of the BISVM is 3.3% and 0.3% higher than the SVM and ISVM respectively. The average training time of the BISVM is shortened from 1.076 s seconds to 0.073 s seconds compared to the ISVM. The BISVM can not only improve the adaptability of a computer to human brain but also lay the foundation to the realization of a fast, real-time and online BCI system.
Key words: brain computer interface (BCI)    batch incremental support vector machine (BISVM)    electroencephalogram (EEG)    classification
出版日期: 2013-09-05
:  TP 391.4  
基金资助:

 国家自然基金项目资助项目（31100709, 60975079）；上海市教育委员会创新资助项目（11YZ19,12ZZ099）.
作者简介: 杨帮华（1971—）,女,副教授,主要从事脑机接口,生物医学信号处理,模式识别与智能系统等研究. E-mail: yangbanghua@shu.edu.cn
服务　　
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章　　

引用本文:

杨帮华, 何美燕, 刘丽, 陆文宇. 脑机接口中基于BISVM的EEG分类[J]. J4, 2013, 47(8): 1431-1436.

YANG Bang-hua, HE Mei-yan, LIU Li, LU Wen-yu. EEG classification based on batch incremental SVM in
brain computer interfaces. J4, 2013, 47(8): 1431-1436.

链接本文:

http://www.zjujournals.com/xueshu/eng/CN/10.3785/j.issn.1008-973X.2013.08.016    或    http://www.zjujournals.com/xueshu/eng/CN/Y2013/V47/I8/1431

［1］ XAVIER A, IMRAN KHAN N, MARIE FRANCOISE L, et al. Accuracy of a BCI based on movement-related and error potentials ［C］∥ 33rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society. Boston: IEEE, 2011: 3688-3691.

［2］ MOLINA G G. BCI Adaptation using incremental SVM learning ［C］∥ 3rd International IEEE/EMBS Conference on Neural Engineering. Hawaii: IEEE, 2007: 337-341.

［3］ ZHANG Ying-chun, HU Guo-sheng, ZHU Feng-Feng, et al. A new incremental learning support vector machine ［C］∥ Artificial Intelligence and Computational Intelligence. Shanghai: Lecture Notes in Computer Science, 2009, 7-10.


［4］ XIAO Huai-tie, SUN Fa-sheng, LIANG Yong-sheng. A fast incremental learning algorithm for SVM Based on K nearest neighbors ［C］∥ Artificial Intelligence and Computational Intelligence. Nanjing: Lecture Notes in Computer Science, 2010, 413-416.

［5］ QIN Jian-zhao, LI Yuan-qing. An improved semi-supervised support vector machine based translation algorithm for BCI systems ［C］∥ The 18th International Conference on Pattern Recognition. Hong Kong: IEEE, 2006, 1240-1243.

［6］ BLANKERTZ B. BCI Competitions. ［EB/OL］. (2008-06-24) ［2011-11-06］. http:∥www.bbci.de/competition

［7］ NOVI Q, GUAN C, DAT T H, et al. Sub-band common spatial pattern (SBCSP) for brain-computer interface ［C］∥ 3rdInternational IEEE/ EMBS Conference on Neural Engineering. Hawaii: IEEE, 2007: 204-207.

［8］ LI Yuan-qing, GUAN Cun-tai. A semi-supervised SVM learning algorithm for joint deature extraction and classification in brain computer interfaces ［C］∥ 28th IEEE EMBS Annual International Conference. New York: IEEE, 2006: 2570-2573.

［9］ MASSIMILIANO P, ALESSANDRO V, ［J］. Neural Computation, 1998, 10(4)： 955-974.

［10］ 吴慧.新的支持向量机增量学习算法［D］.西安：西安电子科技大学, 2009.

WU Hui. New incremental learning algorithms for support vector machines ［D］. Xi’an: Xidian University, 2009.

［11］ ZHU Fa, YE Ning, XU Sheng, et al. Support vectors classification and incremental learning ［C］∥ Information Technology and Artificial Intelligence Conference. Chongqing: IEEE, 2011: 206-210.

［12］ CHRISTOPHER P D, GERT C. SVM Incremental learning, adaptation and optimization ［C］∥ International Symposium on Neural Networks. Oregon : IEEE, 2003: 2685-2690.

［13］ KARASUYAMA, M, TAKEUCHI, I. Multiple incremental decremental learning of support vector Machines ［J］. IEEE Computational Intelligence Society, 2010, 21(7): 1048-1059.
[1] 张林, 程华, 房一泉. 基于卷积神经网络的链接表示及预测方法[J]. 浙江大学学报(工学版), 2018, 52(3): 552-559.
[2] 王卫星, 孙守迁, 李超, 唐智川. 基于卷积神经网络的脑电信号上肢运动意图识别[J]. 浙江大学学报(工学版), 2017, 51(7): 1381-1389.
[3] 杨杨帆, 金平斌, 朱鑫宇. 近30年杭州市城市化进程中土地利用变化[J]. 浙江大学学报(工学版), 2017, 51(7): 1462-1474.
[4] 任迪, 万健, 殷昱煜, 周丽, 高敏. 基于贝叶斯分类的Web服务质量预测方法研究[J]. 浙江大学学报(工学版), 2017, 51(6): 1242-1251.
[5] 朱东阳, 沈静逸, 黄炜平, 梁军. 基于主动学习和加权支持向量机的工业故障识别[J]. 浙江大学学报(工学版), 2017, 51(4): 697-705.
[6] 童基均, 李琳, 林勤光, 朱丹华. 采用平滑伪Wigner-Ville分布的SSVEP脑机接口系统[J]. 浙江大学学报(工学版), 2017, 51(3): 598-604.
[7] 王自立, 张树有, 裘乐淼. 基于可信度区间的注塑装备设计塑化能耗分析[J]. 浙江大学学报(工学版), 2017, 51(2): 328-335.
[8] 邹北骥, 郭建京, 朱承璋, 杨文君, 吴慧, 何骐. BOW-HOG特征图像分类[J]. 浙江大学学报(工学版), 2017, 51(12): 2311-2319.
[9] 李滔, 王士同. 增量式0阶TSK模糊分类器及鲁棒改进[J]. 浙江大学学报(工学版), 2017, 51(10): 1901-1911.
[10] 罗仕鉴, 董烨楠. 面向创意设计的器物知识分类研究[J]. 浙江大学学报(工学版), 2017, 51(1): 113-123.
[11] 冀瑜,邱清盈,冯培恩,黄浩. 国际专利分类表中设计知识的提取和利用[J]. 浙江大学学报(工学版), 2016, 50(3): 412-418.
[12] 裘日辉, 刘康玲, 谭海龙, 梁军. 基于极限学习机的分类算法及在故障识别中的应用[J]. 浙江大学学报(工学版), 2016, 50(10): 1965-1972.
[13] 魏超, 罗森林, 张竞, 潘丽敏. 自编码网络短文本流形表示方法[J]. 浙江大学学报(工学版), 2015, 49(8): 1591-1599.
[14] 谭海龙, 刘康玲, 金鑫, 石向荣, 梁军. 基于μσ-DWC特征和树结构M-SVM的多维时间序列分类[J]. 浙江大学学报(工学版), 2015, 49(6): 1061-1069.
[15] 杨帮华,韩志军,王倩,何亮飞. 分形维数结合RLS-ICA的脑电信号消噪[J]. 浙江大学学报(工学版), 2014, 48(7): 1234-1240.