On loss functions and evaluation metrics for music source separation
This work addresses the problem of evaluating and improving music source separation methods for audio processing researchers, but it is incremental as it focuses on benchmarking existing losses rather than introducing new ones.
The paper benchmarks various loss functions for music source separation to identify which yield better separations, and explores using these losses as evaluation metrics by correlating them with subjective test results, noting that standard metrics can be misleading.
We investigate which loss functions provide better separations via benchmarking an extensive set of those for music source separation. To that end, we first survey the most representative audio source separation losses we identified, to later consistently benchmark them in a controlled experimental setup. We also explore using such losses as evaluation metrics, via cross-correlating them with the results of a subjective test. Based on the observation that the standard signal-to-distortion ratio metric can be misleading in some scenarios, we study alternative evaluation metrics based on the considered losses.