基于状态空间模型的脑电去伪迹与节律提取  

Electroencephalogram artifacts removal and rhythm extraction based on state space model

作　　者：黄丽敏[1] 郝崇清[2] 李斌[3] 

机构地区：[1]河北科技大学现代教育技术中心,河北石家庄050018 [2]河北科技大学电气信息学院,河北石家庄050018 [3]中国民航大学航空自动化学院,天津300300

出　　处：《河北工业科技》2013年第3期147-151,共5页Hebei Journal of Industrial Science and Technology

摘　　要：利用状态空间建模方法实现脑电时间序列的去伪迹与节律提取。通过建立脑电时间序列的自回归滑动平均模型,并将其表示成状态空间并联隔间形式,利用Kalman滤波器对状态进行递推估计实现脑电信号节律提取及去伪迹。给出了单导和多导闭眼静息脑电的两个应用实例,并将多导脑电估计结果和独立成分分析进行了比较。结果表明,该方法能较好地估计出眼动伪迹、工频干扰、alpha节律等成分,且此三种成分与状态转移矩阵的特征值存在对应关系;同时多导脑电估计克服了独立成分的限制,能够得到更多的成分变量。State space model is applied to remove artifacts and extract rhythms from electroencephalogram （EEG） time series. Autoregressive moving average model constructed from EEG is manipulated into a compartment model of the state space. With the state variables estimated recurrently by Kalman filter, artifacts are removed and the rhythm components are extracted. The approach is illustrated by means of two application examples of one dimension and multi--dimension EEG during eye--close resting conditions, and the comparison between the latter example and independent component analysis are made. The results indicate that the two examples successfully realize the blink artifacts, power line interference and alpha rhythm extraction, which are associated with eigenvalues of the state transition matrix. The method used in this paper obtains more components than the independent component analysis.

关 键 词：状态空间模型 节律提取 伪迹消除 独立成分分析 脑电图 KALMAN滤波器 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]