English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Many audio processing tasks require perceptual assessment. The ``goldstandard`` of obtaining human judgments is time-consuming, expensive, andcannot be used as an optimization criterion. On the other hand, automatedmetrics are efficient to compute but often correlate poorly with humanjudgment, particularly for audio differences at the threshold of humandetection. In this work, we construct a metric by fitting a deep neural networkto a new large dataset of crowdsourced human judgments. Subjects are promptedto answer a straightforward, objective question: are two recordings identicalor not? These pairs are algorithmically generated under a variety ofperturbations, including noise, reverb, and compression artifacts; theperturbation space is probed with the goal of efficiently identifying thejust-noticeable difference (JND) level of the subject. We show that theresulting learned metric is well-calibrated with human judgments, outperformingbaseline methods. Since it is a deep network, the metric is differentiable,making it suitable as a loss function for other tasks. Thus, simply replacingan existing loss (e.g., deep feature loss) with our metric yields significantimprovement in a denoising network, as measured by subjective pairwisecomparison.

A Differentiable Perceptual Audio Metric Learned from Just Noticeable Differences