基于信息约简的脑电信号特征提取

Feature extraction of electroencephalogram based on data reduction

背景:脑机接口系统中,对大脑想象的反应速度是衡量脑机接口系统质量的一个重要标准,因此关于脑电信号的分析,对脑电信号特征进行筛选就显得尤为重要,尽可能少的特征来标示相应的动作是提高速度的一个重要途径。目的:利用约简方法对脑电信号特征进行特征筛选,降低脑电特征数目。方法:首先利用不同的脑电特征分析方法,提取特征,并进行分类;其次把连续的脑电信号离散化,利用脑电信号组建脑电信号信息表,最后利用约简理论,对脑电信号进行特征约简,并利用约简特征进行分类,验证分类准确率。结果与结论:利用信息约简的方法及特征标示,把连续的脑电信号特征离散化,利用离散化的脑电信号特征构建信息表,并进行特征筛选,在保证分类准确率的情况下降低了特征的数量。不足之处在于,该方法对分类特征进行标记只能在两类之间进行,对多类型分类特征进行标记和分类特征进行约简将是下一步工作的重点。

BACKGROUND：Reaction velocity for brain imagination is a major criterion in measuring quality of brain computer interface（BCI） system.Therefore,electroencephalogram（EGG） analysis,especially the feature extraction screening,is very important.Reduction the number of features is an important way to improve the speed.OBJECTIVE：To screen EGG features using reduction algorithm,and to decrease the number of EGG features.METHODS：Firstly,by various analytical methods,the EGG features were extracted and classified;then,discreting the continuous EGG to establish an EGG information table;At last,reduction theory was used to reduce EGG features,and to sort the data according to reduced attributes,the accuracy of sorting was validated.RESULTS AND CONCLUSION：Using reduction algorithm and feature marking,discreting the continuous EGG to establish an EGG information table,and then choose the features from the discrete data.The results show that classification accuracy has not been reduced but the number of features was reduced.However,this method can be used marking features in two sorts,how to marking features of multi-sorts and to perform data reduction is a key point in further study.