Spectrogram Feature Losses for Music Source Separation

Venue

arXiv:1901.05061 [cs, eess, stat], vol.

Publication Year

2019

Keywords

Statistics - Machine Learning,Computer Science - Machine Learning,Computer Science - Sound,Electrical Engineering and Systems Science - Audio and Speech Processing,H.5.5,I.2.6,62; 68

Authors

  • Abhimanyu Sahai
  • Romann Weber
  • Brian McWilliams

Abstract

In this paper we study deep learning-based music source separation, and explore using an alternative loss to the standard spectrogram pixel-level L2 loss for model training. Our main contribution is in demonstrating that adding a high-level feature loss term, extracted from the spectrograms using a VGG net, can improve separation quality vis-a-vis a pure pixel-level loss. We show this improvement in the context of the MMDenseNet, a State-of-the-Art deep learning model for this task, for the extraction of drums and vocal sounds from songs in the musdb18 database, covering a broad range of western music genres. We believe that this finding can be generalized and applied to broader machine learning-based systems in the audio domain.