Film Affective Content Recognition Based on Fuzzy Inference

Affective content plays an important role in film analysis and retrieval. However, the widely affective gap between the low-level features and the emotion recognition is still an unsolved problem. In order to recognize the affective type of a scene, a new algorithm is proposed based on fuzzy inference theory in this paper. It contains three main technologies. Firstly, two feature extraction models are built up by analyzing the film grammar. Secondly, based on fuzzy membership functions, a fuzzy method is used to transform a low-level feature vector into a fuzzy vector. This fuzzy expression of the scene content is more closed to the humanpsilas emotion description. Thirdly, a fuzzy logic inference rules are established to infer the affective type of a scene based on self-assessment report and fuzzy theory. Experimental results show that the proposed algorithm is feasible and achieves a high recognition accuracy which exceeds 80 percent.