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 our previous work, we proposed a neural vocoder called APNet, whichdirectly predicts speech amplitude and phase spectra with a 5 ms frame shift inparallel from the input acoustic features, and then reconstructs the 16 kHzspeech waveform using inverse short-time Fourier transform (ISTFT). APNetdemonstrates the capability to generate synthesized speech of comparablequality to the HiFi-GAN vocoder but with a considerably improved inferencespeed. However, the performance of the APNet vocoder is constrained by thewaveform sampling rate and spectral frame shift, limiting its practicality forhigh-quality speech synthesis. Therefore, this paper proposes an improvediteration of APNet, named APNet2. The proposed APNet2 vocoder adopts ConvNeXtv2 as the backbone network for amplitude and phase predictions, expecting toenhance the modeling capability. Additionally, we introduce a multi-resolutiondiscriminator (MRD) into the GAN-based losses and optimize the form of certainlosses. At a common configuration with a waveform sampling rate of 22.05 kHzand spectral frame shift of 256 points (i.e., approximately 11.6ms), ourproposed APNet2 vocoder outperformed the original APNet and Vocos vocoders interms of synthesized speech quality. The synthesized speech quality of APNet2is also comparable to that of HiFi-GAN and iSTFTNet, while offering asignificantly faster inference speed.

APNet2: High-quality and High-efficiency Neural Vocoder with Direct Prediction of Amplitude and Phase Spectra