Affective Classification Method Based on Movie 5.1 Sound
He Mu, Yang Yong-en, Luo Chao-liang · 2020
In this paper, a method using the 5.1 surround sound format to classify the affective scenes of a movie was introduced. The affective types were decided by a web survey and several subjective listening tests, then a movie scene library with labeled affective types was obtained. The content and characteristic of the 5.1 surround sound format was analyzed, and feature parameters of the main left channel and the low-frequency effect channel were extracted as the materials for the prediction calculation. Several classic models have been used and the prediction results were compared among five frame lengths of the audio signals, and 2s used Bagged Trees model got the highest predicting accuracy.