MGEPM: Specific Composition Style Melody Generation Based on Music Gene Expression Programming

This study proposes a music gene expression pro-gramming method to achieve music melody feature learning of a specific singer's compositional style. The user specifies the song format, and the program automatically generates and evaluates songs. A mapping mechanism between 16-bit octal numbers and musical scores is proposed to realize the evolution of musical scores in units of measures. In the melody generation stage, the modification operations that are generally performed by the composer in the process of composing music are integrated with the evolutionary operations of gene expression programming to create a music gene expression programming operation. In the music melody evaluation stage, a long short-term memory neural network evaluation model is established and trained using the main verse and chorus of the music as well as the spectrum of the entire song. The neural network evaluates the resulting song fitness value, providing feedback for music evolution. In contrast to melody generated via other methods, comparative experiments show that the melody generated by this research method has an obvious compositional style of a specific composer.