Testing a new methodology for accelerating the computation of quadratic sample entropy in emotion recognition systems

Abstract

Emotion recognition has become an important area of study for the development of human-machine interfaces able to recognize and interpret human emotions. In order to construct emotional systems, signals from physiological variables have to be registered and processed rapidly to provide a fast emotional response from the computer system to the user. In this regard, several studies have claimed that nonlinear methodologies applied to electroencephalographic signals can provide relevant information about emotions recognition. However, given the multimodal nature and nonlinear behaviour of that signals, the data processing is often very slow to give a fast response, producing an important delay between feeling an emotion and receiving the adequate response from the emotional system. In order to overcome this difficulty, this work computes a modification of quadratic sample entropy accelerating the computation by exploiting vectors with dissimilarity.