Based on Multiscale Permutation Entropy Analysis Dynamic Characteristics of EEG Recordings

In this study, a novel feature extraction called entropy is proposed for evaluating the dynamic characteristics of electroencephalogram (EEG). Meanwhile, the linear discriminant analysis(LDA) is used in the classification of EEG signals. The entropy values are calculated using the multiscale permutation entropy (MPE) method, in order to evaluate their performances in terms of classification accuracy, the experiment is implemented using different EEG signals. When all EEGs can be divided into three groups of healthy segments, epileptic seizure-free segments, and epileptic seizure segments, the classification accuracy is as high as 93%. The experiments show that the LDA-based multiscale permutation entropy approach achieves higher classification accuracy rates than the that of single-scale permutation entropy (PE) approach, and MPE can reveal the hidden characteristics of EEG signals.