Research on Chord-Constrained Two-Track Music Generation Based on Improved GAN Networks

Chords have a role in music for emotional expression, and the generated melodies more richness through constraining effect of chords. In this paper, based on GAN network generation model chord features, GRU is used feature extraction order to autonomously learn chords at 1 : t − moments generate moments, by saving hidden layer state each batch constructing combined with generator, thus achieving automatically learning overall style The performance four models gradually optimized weighted averaging, melodic pleasantness all has significant positive correlation musical coherence creativity.