Application of EEMD-HHT Method on EEG Analysis for Speech Evoked Emotion Recognition

Abstract

Electroencephalograph (EEG) is widely used to study human brain activities. However, the interpretation of EEG signals is still a challenging computational task. Emerging evidence has shown that the non-stationary traits of EEG signals hinder the way of informative interpretation. Compared to the classical Welch frequency analysis method (short-time Fourier transform), Hilbert Huang Transform(HHT) is more suitable for non-linear and non-stationary signals. This paper proposes a band energy extraction method based on EEMD-HHT for time-frequency analysis of EEG signals. We evaluate the method on an EEG database obtained through the emotional cognitive experiment. The auditory stimulus in this paper are selected from CHEAVD2 which is a speech emotion database of the Chinese Academy of Sciences. The correlation coefficients between the predict and target values reach 0.51 and 0.43 for arousal and valence dimension, respectively. This method shows great potentials in applications of computational neuroscience and cognition of art creation.