Non-lineal EEG modelling by using quadratic entropy for arousal level classification

Abstract

Nowadays, assistive technologies together with ubiquitous and pervasive computing are emerging as main alternative to help ageing population. In this respect, an important number ofworks have been carried out to improve the quality of life in elderly from a physical point of view. However, less efforts have been made in monitoring the mental and emotional states of the elderly. This work presents a nonlinear model for discriminating different arousal levels through quadratic entropy and a decision tree-based algorithm. Two hundred and seventy eight EEG recordings lasting one minute each were used to train the proposed model. The recordings belong to the Dataset for Emotion Analysis using Physiological signals (DEAP). In agreement with the complexity and variability observed in other works, our results report a low quadratic entropy when subjects face high arousal stimuli. Finally, the model achieves a global performance around 70% when discriminating between calm and excitement events.