Comparing the effect of audio coding artifacts on objective quality measures and on subjective ratings

A recent work presented the subjective ratings from an extensive perceptual quality evaluation of audio signals, where isolated coding artifact types of varying strength were introduced. We use these ratings as perceptual reference for studying the performance of 11 well-known tools for objective audio quality evaluation: PEAQ, PEMO-Q, ViSQOLAudio, HAAQI, PESQ, POLQA, fwSNRseg, dLLR, LKR, BSSEval, and PEASS. Some tools achieve high correlation with subjective data for specific artifact types (Pearson's r > 0.90, Kendall's t > 0.70), corroborating their value during the development of a specific algorithm. Still, the performance of each tool varies depending on the artifact type and no tool reliably assesses artifacts from parametric audio coding. Nowadays, perceptual evaluation remains irreplaceable, especially when comparing different coding schemes introducing different artifacts.