Weighted-permutation entropy as complexity measure for electroencephalographic time series of different physiological states

An electroencephalographic (EEG) waveform could be denoted by a series of ordinal patterns called motifs which are based on the ranking values of subsequence time series. Permutation entropy (PE) has been developed to describe the relative occurrence of each of these motifs. However, PE has few limitations, mainly its inability to differentiate between distinct patterns of a certain motif, and its sensitivity to noise. To minimize those limitations, Weighted-Permutation Entropy (WPE) was proposed as a modification version of PE to improve complexity measuring for times series. This paper presents an approach by incorporating WPE into the analysis of different physiological states namely EEG time series. Three different EEG physiological states, eye-closed (EC), eye-open (EO), and visual oddball task (VOT) were included to examine ability of WPE to identify and discriminate different physiological states. The classification using WPE has achieved the results with accuracy of 87% between EC and EO states, and 83% between EO and VOT, respectively, using linear discrimination analysis. The results showed the potential of WPE to be a promising feature for nonlinear analysis in different physiological states of brain. It was also observed that WPE also could be used as marker for large artifact with low frequency such as eye-blink.