eduzhai > Applied Sciences > Engineering >

Transformer-XL Based Music Generation with Multiple Sequences of Time-valued Notes

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 9 pages

Abstract: Current state-of-the-art AI based classical music creation algorithms such asMusic Transformer are trained by employing single sequence of notes withtime-shifts. The major drawback of absolute time interval expression is thedifficulty of similarity computing of notes that share the same note value yetdifferent tempos, in one or among MIDI files. In addition, the usage of singlesequence restricts the model to separately and effectively learn musicinformation such as harmony and rhythm. In this paper, we propose a frameworkwith two novel methods to respectively track these two shortages, one is theconstruction of time-valued note sequences that liberate note values fromtempos and the other is the separated usage of four sequences, namely, formernote on to current note on, note on to note off, pitch, and velocity, forjointly training of four Transformer-XL networks. Through training on a 23-hourpiano MIDI dataset, our framework generates significantly better and hour-levellonger music than three state-of-the-art baselines, namely Music Transformer,DeepJ, and single sequence-based Transformer-XL, evaluated automatically andmanually.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...