A novel affect-based model of similarity measure of videos

With the fast proliferation of multimedia and video display devices, searching and watching videos on the Internet has become an indispensable part of our daily lives. Many video-sharing web sites offer the service of searching and recommending videos from an exponentially growing repository of videos uploaded by individual users. The issue of finding videos suited to a user's personal preferences or measuring the similarity between videos poses various challenges. In this paper, we present a novel affect-based model of similarity measure of Internet videos, which can be used in the video retrieval and recommendation systems. We employ our similarity measure model on normalized V–A (valence–arousal) graphs formed by extracting four basic affective video features, i.e., motion, shot-change rate, sound energy and audio pitch average. Our experiments, based on qualitative comparison of our method with the Affivir method for measuring affect-based video similarity, demonstrate the superiority of our proposed model.