Audience Oriented Personalized Movie Affective Content Representation and Recognition

In order to represent and recognize the movie affective content reasonably and automatically, an audience oriented personalized movie emotion space modeling method is proposed as a solution for the "affective gap".It is an essential problem on the movie affective semantic understanding.By our method, a fuzzy c-mean clustering (FCM) algorithm is adopted to divide the Valence-Arousal emotion space into the typical fuzzy emotion subspaces and Gaussian mixture model (GMM) is used to determine their affective membership functions.The centers, borders, shapes and densities of these subspaces can truthfully reflect the emotional tendencies of audiences.Two sets of movie affective feature vectors are formulated.Multi-layer perception (MLP) and multiple linear regression are adopted to compute the emotion coordinates of these movie affective feature vectors.Based on the affective membership functions and emotion coordinates, the maximum membership principle and the threshold principle are introduced to represent and recognize the emotional preferences of the audiences.Experimental results demonstrate that the proposed modeling method can be applied to effectively represent and recognize the personalized movie affective content.