基于扩展Infomax ICA的ERP少次提取方法研究

事件相关电位(event related potential,ERP)提取是脑电研究的重点之一,目前临床上主要通过相干平均的方法来获取.由于脑电的非平稳性,使其需要大量重复刺激才能获得,对于受试者极不方便,也不利于ERP的实时检测.本文以反映大脑稀少认知事件的相关电位P300为例,采用扩展Infomax(extended informationmaximization)独立分量分析(independent component analysis,ICA)算法,先滤除眼动、工频干扰,再重构脑电数据,最后经少次叠加即可得到与通常需多次相干平均结果相近的比较满意的P300波形.说明ICA算法在ERP的峰值和潜伏期模式识别上具有较为明显的效果,具有潜在的临床工程应用价值.

Study on the Several Trials Extraction of ERP Based on Extended-Infomax ICA

The extraction of Event Related Potential (ERP) is one of important points of EEG research. And it is obtained by the method of coherent averaging at the present clinical study. But this way is extremely inconvenient to the test receivers because the EEG signal is non-stationary so that it needs the same stimulations at least over hundred times, furthermore, it has difficulty for the estimation of ERP on time. This paper lists the examples of P300 that reflects the related potential of the brain's recognizing rare events. A method based on the extended infomax of independent component analysis (ICA) is proposed for seldom trials extraction of ERP. In this algorithm, firstly the interferences such as blink artifacts and power noise are removed, then the EEG data is rebuilt finally the several trails are added in order to get a rather satisfied P300 wave patter which is similar to that obtained by the conventional coherent averaging needed over 36 times. It illustrates that the ICA algorithm has an obvious effect on the pattern recognition of ERP peaks & latencies, and is valuable for application in clinical engineering.