Movie Genre Classification Using SVM with Audio and Video Features

摘要

In this paper, we propose a movie genre classification system using a meta-heuristic optimization algorithm called Self-Adaptive Harmony Search (i.e., SAHS) to select local features for corresponding movie genres. Then, each one-against-one Support Vector Machine (i.e., SVM) classifier is fed with the corresponding local feature set and the majority voting method is used to determine the prediction of each movie. Totally, we extract 277 features from each movie trailer, including visual and audio features. However, no more than 25 features are used to discriminate each pair of movie genres. The experimental results show that the overall accuracy reaches 91.9%, and this demonstrates more precise features can be selected for each pair of genres to get better classification results.