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/

Global: Zendy Plus annual

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

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In deep learning research, many melody extraction models rely on redesigningneural network architectures to improve performance. In this paper, we proposean input feature modification and a training objective modification based ontwo assumptions. First, harmonics in the spectrograms of audio data decayrapidly along the frequency axis. To enhance the model's sensitivity on thetrailing harmonics, we modify the Combined Frequency and Periodicity (CFP)representation using discrete z-transform. Second, the vocal and non-vocalsegments with extremely short duration are uncommon. To ensure a more stablemelody contour, we design a differentiable loss function that prevents themodel from predicting such segments. We apply these modifications to severalmodels, including MSNet, FTANet, and a newly introduced model, PianoNet,modified from a piano transcription network. Our experimental resultsdemonstrate that the proposed modifications are empirically effective forsinging melody extraction.

Towards Improving Harmonic Sensitivity and Prediction Stability for Singing Melody Extraction