Analysis of Guzheng Music Style Transformation Based on Generative Adversarial Networks

Abstract

With the emergence of more and more new music styles, Guzheng, a traditional national musical instrument that has been handed down for thousands of years, also faces new challenges. However, with the development of the times and aesthetics, composers have created various modern Guzheng music, thus injecting new vitality into the development of Guzheng. In order to study the mutual conversion between Guzheng music and other music, this paper proposes a Guzheng music conversion model based on Star generative adversarial networks (GAN), which can convert any length of music and realize the music conversion between various styles. Based on the idea of star GAN, this model uses Mel spectrum of music and random target style tags to train a general generator to generate a specified music style. It is found that the model proposed in this paper has better conversion effect between similar styles of music; that is, when Guzheng music is converted into pop and flute, the effect is the best, the average accuracy is 89.62% and 90.32%, respectively, and the cosine similarity is also 15.8 and 20.5.