A biologically inspired model of emotion eliciting from visual stimuli

Abstract

Several emotion eliciting models have been proposed in the literature, however most of them are still artificial models which ignore the biological basis. We propose an emotion (without awareness) eliciting model from visual stimuli, which is inspired by biology: we describe an emotion eliciting process that follows the circuits of emotion in the brain derived through the results of neuroscience and the three major modules in the process, visual perception, emotion-eliciting region and emotional valence elicited by the region, are all supported by biology research. In our work, visual perception works with visual stimuli from coarse to the finer level according to human visual system. The elicited emotion in coarse level is also capable of affecting the emotion valence in the finer level. Based on psychophysical research, the emotion-eliciting region is selected out through color preference. The emotion is elicited by the emotion-eliciting region rather than overall visual context, which has been first introduced to computational modeling of emotion eliciting from image stimuli. The emotional valence elicited by the region is calculated on coarseness and directionality by comparing with stored image representations. In the experiments, two types of visual stimuli are considered: (1) natural scenes stimuli and (2) natural scenes and mutilation scenes stimuli. We compare the performance of our model with International Affective Picture System (IAPS), a large set of emotionally evocative color photographs that includes pleasure and arousal ratings made by men and women. Experimental results show that our model can generate human-like emotion based on natural scenes stimuli and obtain the positive or negative emotion as people feel on natural scenes and mutilation scenes stimuli.