A Novel Probabilistic Approach to Modeling the Pleasure-Arousal-Dominance Content of the Video based on "Working Memory"

Abstract

The rapid growth of digital video information nowadays is making video content classification and indexing tools a necessity. Little research efforts have been invested on content classification solutions based on the emotional content of the video, which have the potential of extending the scope of video indexing possibilities. The development of an affective content classification solution faces a number of challenges. These include the dynamic and time evolving nature of the video's emotional content and the uncertainty in the multimodal sensory observations. These challenges have resulted in the lack of reliable methods for video emotional content modeling. This paper introduce a novel probabilistic approach to model the emotional content of the video based on the dynamic Bayesian networks (DBNs). It is designed based on the pleasure-arousal-dominance (P-A-D) emotion model, which in principle can represent a large number of emotions. It is also designed based on the concept of "working memory", a theoretical framework within cognitive psychology that refers to the structures and processes used for temporarily storing and manipulating information. Our experiment results demonstrate that "working memory" is an important aiding factor in complex emotional content analysis.