Nonlinear symbolic assessment of electroencephalographic recordings for negative stress recognition

Abstract

Nowadays, electroencephalographic (EEG) recordings receive increasing attention in the field of emotions recognition with physiological variables. Moreover, the nonlinear nature of EEG signals suggests that nonlinear techniques could be more suitable than linear methodologies for the assessment of mental processes triggered under different emotions. One of the most relevant states is distress (the negative aspect of stress), because of its enormous influence in developed countries and its countless adverse effects in health. As a result, many researches have shown their interest in distress in the last few years. In the present study, a predictability-based entropy measure called amplitude-aware permutation entropy (AAPE) was applied to discern between calm and distress states. EEG signals from 32 channels were individually assessed to obtain the discriminatory ability of each single electrode. After that, only 2 out of 32 EEG channels were combined in a logistic regression model, reaching a global classification accuracy over 73%.