Multimodal Recognition of Freezing Behavior in Virtual Reality

Abstract

Freezing is a component of fear behavior and is an evolutionary defense mechanism found in humans and animals that can be characterized by reduced body motion, decreased heart rate, and increased muscle contractions. In previous literature, fear stimuli were often presented using pictorial stimuli, and behavior was measured intrusively. These factors could negatively influence to what extent natural behavior is shown. Therefore, in this thesis, a novel and less intrusive multimodal approach for recognizing freezing behavior is proposed, which uses virtual reality (VR) to present the stimuli and uses balance and gaze data for the recognition of fear behavior. In VR, the participants were approached by armed and unarmed characters, which were the fear-inducing and neutral stimuli, respectively.

Results of trials featuring a single type of stimuli showed that participants' gaze fixated more on fear-inducing elements of the stimuli, while gaze was more exploratory when neutral stimuli were shown. Additionally, when showing both types of stimuli simultaneously, the gaze fixated more on the fear-inducing stimuli than the neutral stimuli, which is in line with previous literature about fear and freezing behavior. When measuring body sway, it was found that some participants showed freezing behavior, whereas other participants showed opposing behavior. During the fear-inducing trials, adjustments in behavior were found once the stimuli were nearing the participants, suggesting freezing required a delay to manifest in the participants, though evidence of this was not strong.

These results suggest that the proposed method could be used to detect fear and freezing behavior and could lay the foundation for further behavioral research in freezing behavior. Additionally, gaze fixations appeared to be a good measurement of fear behavior.