Discrimination of mental tasks based on EEMD and information theoretic pattern selection

Abstract

In this paper, we address the discrimination of mental tasks problem and suggest a method based on Ensemble Empirical Mode Decomposition (EEMD), for time-frequency analysis, and a pattern selection method based on an information theoretic measure, namely; Jensen Shannon Divergence (JSD) measure. The method works in three steps: (i) to employ EEMD for EEG signal decomposition into components called Intrinsic Mode Functions (IMFs), followed by applying Hilbert transform to the IMFs to determine the instantaneous frequency and amplitude; (ii) to choose the IMFs containing the most significant information based on the degree of presence in gamma band; (iii) to select segments of instantaneous vectors according to JSD metric, which measures the distances between two concepts. This method was applied to EEG signals of 5 subjects performing 5 mental tasks. The classification of mental tasks was performed using Fisher linear discriminator. The experimental results are compared with the ones obtained by a method that uses the power of gamma band in EEG signals (a traditional and popular method). The experimental results show improvement of the classification accuracy.