基于ECoG的运动想象分类

被引:3
作者
安滨 [1 ]
江朝晖 [1 ]
宁艳 [1 ]
陈强 [2 ]
冯焕清 [1 ]
机构
[1] 中国科学技术大学电子科学与技术系
[2] 合肥工业大学生物医学工程系
关键词
运动想象; 皮层脑电图; μ节律; 共同空间特征法滤波; K近邻;
D O I
暂无
中图分类号
R319 [其他科学技术在医学上的应用];
学科分类号
1001 ;
摘要
目的以两种运动想象任务下采集的64导ECoG信号为训练样本,识别几天后重复进行的运动想象任务。方法以动作感知皮层区脑电图(ECoG)的μ节律(8Hz13Hz频段)功率谱为特征。通过手工比较功率谱的差异显著性,从64导中粗选出11导最明显的信号。再用共同空间特征法(CSP)滤波提高信噪比,使信号从11维降到8维。采用K近邻分类器进行分类识别,其中依据交叉验证法得到最佳的近邻值。结果测试样本的预测精度达到94%。结论利用动作感知皮层区脑电μ节律能较好识别对应的特定(想象)运动;共同空间特征法滤波可以有效提高信噪比;只要预处理、特征抽取及分类得当,时间间隔和实验误差等因素对运动想象识别的影响不大。
引用
收藏
页码:64 / 68
页数:5
相关论文
共 12 条
[1]  
Brain-ComputerInterface Research at the Wadsworth Center. Wolpaw JR,McFarland DJ,Vaughan TM. IEEE Trans RehabEng . 2000
[2]  
Brain-computerinterface(BCI)operation:signal and noise during early trainingsessions. McFarland DJ,Sarnacki WA,Vaughan TM,et al. Clinical Neurophysiology . 2005
[3]  
A Direct BrainInterface Based on Event-Related Potentials. Levine SP,Huggins JE,BeMent SL,et al. IEEE Trans RehabEng . 2000
[4]  
Wave number power spectrum of theEED,ECoG,and ERP. O′Connor SC,Robinson PA. Neurocomputing . 2004
[5]  
Mu rhythm-based cursorcontrol:an offline analysis. Cheng M,Jia WY,Gao XR,et al. Clinical Neurophysiology . 2004
[6]  
Classifying EEG-based motor imagerytasks by means of time-frequency synthesized spatial patterns. Wang T,Deng J,He B. Clinical Neurophysiology . 2004
[7]  
Motor-cortical activityin tetraplegics. Shoham S,Halgren E,Maynard EM,et al. Brief Communications Nature . 2001
[8]  
IEEE BCI Competition III. Blankertz B. http://ida.first.fraunhofer.de/projects/bci/competition-iii,2005-07-16/ . 2005
[9]  
Optimal SpatialFiltering of Single Trial EEG During Imagined Hand Movement. Ramoser H,Mlüler-Gerking J,Pfurtscheller G. IEEE Transactions on Rehabilitation Engineering . 2000
[10]  
Learning to control brain activity:A reviewof the production and control of EEG components for driving brain-computer interface(BCI)systems. Curran EA,Stokes MJ. Brain and Cognition . 2003