Research on intelligent algorithm for optimizing the matching between background music and sound effect of film and television works by generating antagonistic network

Abstract

In this paper, an intelligent algorithm based on Generating Antagonistic Network (GAN) is proposed to optimize the matching between background music and sound effects of film and television works. Traditional matching methods rely on manual editing, which is inefficient and difficult to ensure the effect. GAN, as an advanced deep learning model, can generate high-quality and diverse audio data through confrontation training between generator and discriminator. The algorithm framework designed in this study extracts film and television features through feature fusion encoder, and uses generator to generate matching audio, while discriminator evaluates the authenticity of generated audio. In addition, the algorithm introduces emotional consistency loss and multi-scale spectrum matching loss to enhance the emotional expression and sound-picture matching degree of generated audio. The experimental results show that this method is superior to the existing methods in both objective and subjective indexes, and the accuracy of emotion matching is over 80%, and the training efficiency and style diversity are excellent. This study provides an efficient and convenient solution for the automatic and intelligent matching of background music and sound effects in film and television works, which has important application value.