基于多尺度模糊熵的脑电信号特征提取方法

EEG Feature Extraction Method Based on Multiscale Fuzzy Entropy

目的利用多尺度模糊熵特征提高脑电信号的识别率。方法采用滑动均值滤波使粗粒化后脑电信号各尺度上的序列尽量保留原始信息,然后用多尺度模糊熵进行特征提取,通过Fisher得分对不同尺度模糊熵的特征分类能力排序,选定排序靠前的多尺度模糊熵组成最优特征向量,实现特征降维。结果在BCI Competition实验数据上进行验证,结果表明该方法用支持向量机做分类器,准确率平均达88.94%。结论本文提出的多尺度模糊熵对脑电信号的特征提取方法具有针对性,既提升了识别率又保持了算法效率。

Objective To improve the recognition rate of EEG signals by multiscale fuzzy entropy. Methods A moving-averaging procedure was used to obtain the coarse-grained time series which mostly kept the original information of EEG signals. Then the features were extracted by multiscale fuzzy entropy. The higher-ranking multiscale fuzzy entropy,whose feature classification ability of different scales were sorted by Fisher,was selected as the optimal eigenvector and the feature dimension reduction was realized. Results The data sets of BCI Competition were used to verify the method. The results showed that the classification rate reached 88.94% on average when applied to support vector machine as a classifier. Conclusion The multiscale fuzzy entropy proposed in this paper is suitable for EEG extraction,which not only improves the recognition rate but also maintains the efficiency of the algorithm.