DCGAN-BASED MUSIC GENERATION METHOD, AND MUSIC GENERATION APPARATUS

摘要

Provided are a DCGAN-based music generation method, and a music generation apparatus. The method comprises: constructing a deep convolution generative adversarial network (DCGAN) model (S101); then acquiring a training data set (S102); next, inputting N melody matrices and N corresponding chord matrices in the training data set into the DCGAN model for training so as to obtain a trained DCGAN model (S103); and then inputting an acquired target melody matrix into the trained DCGAN model for processing, acquiring a target chord matrix generated by the trained DCGAN model and matching the target melody matrix, and finally, outputting a music file formed after combining a melody track mapped by the target melody matrix with a chord track mapped by the target chord matrix (S104). By means of the music generation method, a music file with matching chords can be automatically generated, thereby reducing manual handling steps.