A stochastic process model of melody generation in popular music composition and its contribution to compositional innovation

Digital generation of musical melodies is a new field in improving popular music composition, but due to the stochastic nature of music composition, it is necessary to model and analyze the stochastic process of melody generation. Melody generation is modeled using a classical Markov chain, based on which the Markov algorithm is improved by adding constraints to combine the generated melody and rhythm. The subjective score performance of the generated melodies was verified using three sampling methods, and it was found that the melodies generated by the present method improved their subjective scores by about 21.95%, 31.88%, and 30% compared to those of the traditional Markov. In terms of phrase relevance, interval characteristics, and number of short rhythms, the overall melodic performance of this method is about 1~1.7 times higher than that of traditional Markov and attentional_rnn models. It shows that the method in this paper can indeed generate high-quality melodies for popular music and provide impetus for compositional innovation.