PrincetonComputer SciencePIXL GroupPublications → [Manocha et al. 2020] Local Access
A Differentiable Perceptual Audio Metric Learned from Just Noticeable Differences

arXiv, January 2020

Pranay Manocha, Adam Finkelstein, Zeyu Jin,
Nicholas J. Bryan, Richard Zhang, Gautham J. Mysore
Reference Audio1 Audio2
Abstract

Assessment of many audio processing tasks relies on subjective evaluation which is time-consuming and expensive. Efforts have been made to create objective metrics but existing ones correlate poorly with human judgment. In this work, we construct a differentiable metric by fitting a deep neural network on a newly collected dataset of just-noticeable differences (JND), in which humans annotate whether a pair of audio clips are identical or not. By varying the type of differences, including noise, reverb, and compression artifacts, we are able to learn a metric that is well-calibrated with human judgments. Furthermore, we evaluate this metric by training a neural network, using the metric as a loss function. We find that simply replacing an existing loss with our metric yields significant improvement in denoising as measured by subjective pairwise comparison.
Additional Links
Citation

Pranay Manocha, Adam Finkelstein, Zeyu Jin, Nicholas J. Bryan, Richard Zhang, and Gautham J. Mysore.
"A Differentiable Perceptual Audio Metric Learned from Just Noticeable Differences."
arXiv, January 2020.

BibTeX

@inproceedings{Manocha:2020:ADP,
   author = "Pranay Manocha and Adam Finkelstein and Zeyu Jin and Nicholas J. Bryan
      and Richard Zhang and Gautham J. Mysore",
   title = "A Differentiable Perceptual Audio Metric Learned from Just Noticeable
      Differences",
   booktitle = "arXiv",
   year = "2020",
   month = jan
}