Musical Source Separation Bake-Off: Comparing Objective Metrics with Human Perception

TL;DR

This study evaluates how well various objective metrics predict human perception of music source separation quality, revealing that no single metric is universally reliable and emphasizing the importance of stem-specific evaluation methods.

Contribution

It provides a large-scale listener dataset and compares multiple objective metrics, highlighting their strengths and limitations in predicting human perception across different music stems.

Findings

SDR best predicts vocal quality

SI-SAR better correlates with perception for drums and bass

FAD with CLAP-LAION-music performs well for drums and bass

Abstract

Music source separation aims to extract individual sound sources (e.g., vocals, drums, guitar) from a mixed music recording. However, evaluating the quality of separated audio remains challenging, as commonly used metrics like the source-to-distortion ratio (SDR) do not always align with human perception. In this study, we conducted a large-scale listener evaluation on the MUSDB18 test set, collecting approximately 30 ratings per track from seven distinct listener groups. We compared several objective energy-ratio metrics, including legacy measures (BSSEval v4, SI-SDR variants), and embedding-based alternatives (Frechet Audio Distance using CLAP-LAION-music, EnCodec, VGGish, Wave2Vec2, and HuBERT). While SDR remains the best-performing metric for vocal estimates, our results show that the scale-invariant signal-to-artifacts ratio (SI-SAR) better predicts listener ratings for drums and bass stems. Frechet Audio Distance (FAD) computed with the CLAP-LAION-music embedding also performs competitively--achieving Kendall's tau values of 0.25 for drums and 0.19 for bass--matching or surpassing energy-based metrics for those stems. However, none of the embedding-based metrics, including CLAP, correlate positively with human perception for vocal estimates. These findings highlight the need for stem-specific evaluation strategies and suggest that no single metric reliably reflects perceptual quality across all source types. We release our raw listener ratings to support reproducibility and further research.