Emotion Pattern Recognition Using Physiological Signals

Abstract

In this paper, we first regard emotion recognition as a pattern recognition problem, a novel feature selection method was presented to recognize human emotional state from four physiological signals. Electrocardiogram (ECG), electromyogram (EMG), skin conductance (SC) and respiration (RSP). The raw training data was collected from four sensors, ECG, EMG, SC, RSP, when a single subject intentionally expressed four different affective states, joy, anger, sadness, pleasure. The total 193 features were extracted for the recognition. A music induction method was used to elicit natural emotional reactions from the subject, after calculating a sufficient amount of features from the raw signals, the genetic algorithm and the K-neighbor methods were tested to extract a new feature set consisting of the most significant features which represents exactly the relevant emotional state for improving classification performance. The numerical results demonstrate that there is significant information in physiological signals for recognizing the affective state. It also turned out that it was much easier to separate emotions along the arousal axis than along the valence axis.