基于振动模态参数识别的脑电信号特征提取

Feature Extraction of EEG Signal Based on Recognition of Vibration Modal Parameters

对运动想象脑电信号的动力学模型进行了分析,将其分成两个阶段(强非线性的瞬态阶段和弱非线性的自由响应阶段),并构建了一种新的特征提取算法。首先通过起始点扫描的方式对脑电信号进行分割来获得自由响应阶段的脑电信号;然后针对自由响应阶段产生的脑电信号,引入振动多模态参数识别ITD(Ibrahim Time Domain)算法来提取特征组合成特征向量;最后利用Adaboost分类器进行自适应特征选择和分类。运用此方法对国际标准数据库The largest SCP data of Motor-Imagery中的CLA运动想象数据集进行特征提取和特征选择与分类,其平均分类准确率高达90%以上。与现有的特征提取算法相比,获得了更好的分类性能和稳定性。

The dynamic model of motor imagination EEG is analyzed and divided into two stages:a transient phase with strong nonlinearity and a free response phase with weak nonlinearity.A new experimental method of EEG feature extraction is proposed in this paper.The EEG signal is segmented by scanning the starting point to obtain the EEG signal in the free response phase firstly.Subsequently,for the EEG signal generated during the free-response vibration phase,the vibration multi-modal parameter recognition ITD(Ibrahim Time Domain)algorithm is introduced to extract features and combine them into feature vectors.Finally,the obtained feature vector is used for adaptive feature selection and classification using Adaboost classifier.This method is used to extract,select and classify features from CLA Motor Imagery dataset in The largest SCP Data of Motor Imagery.The average classification accuracy rate is as high as 90%.Compared with the existing feature extraction algorithms,the classification performance and stability are better.