Recognition of Distress Phase Situation in Human Emotion EEG Physiological Signals

Abstract

Emotions play a vital role in humans’ daily life and aid in understanding our social behaviors. They are richly embedded in human nonverbal communications and can be creatively utilized to prompt an emergency assistance in a distress phase situation to enhance human safety, thus making human emotion recognition a perplexing task. In this study, we constructed a distress phase emotion model using the inherent properties of the arousal, valence, dominance and liking dimensions of emotion representation. Furthermore, we introduce a shifted tanh-based normalization scheme and apply the inverse Fisher transformation algorithm to the Database for Emotion Analysis using Physiological Signal (DEAP) dataset. The Radial Basis Function Artificial Neural Networks (RBF-ANN) pattern recognizer was consequently utilized to conduct various experiments, and the performances of digital image-based feature extraction techniques of Histogram of Oriented Gradient (HOG), Local Binary Pattern (LBP) and Histogram of Images (HIM) are compared. The experimental results obtained indicate that the best recognition accuracy of 91.41% was achieved with the HIM features extracted from the electroencephalogram (EEG) data of the DEAP dataset with classification done along the tripartite classes of happy, distress and casualty of our constructed emotion model. The results obtained are very remarkable when compared with the existing results in the literature including some deep learning studies that have utilized the same DEAP corpus, and thus are applicable for prompting emergency services in distress phase situations.