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

Recent development of neural vocoders based on the generative adversarialneural network (GAN) has shown obvious advantages of generating raw waveformconditioned on mel-spectrogram with fast inference speed and lightweightnetworks. Whereas, it is still challenging to train a universal neural vocoderthat can synthesize high-fidelity speech from various scenarios with unseenspeakers, languages, and speaking styles. In this paper, we propose DSPGAN, aGAN-based universal vocoder for high-fidelity speech synthesis by applying thetime-frequency domain supervision from digital signal processing (DSP). Toeliminate the mismatch problem caused by the ground-truth spectrograms in thetraining phase and the predicted spectrograms in the inference phase, weleverage the mel-spectrogram extracted from the waveform generated by a DSPmodule, rather than the predicted mel-spectrogram from the Text-to-Speech (TTS)acoustic model, as the time-frequency domain supervision to the GAN-basedvocoder. We also utilize sine excitation as the time-domain supervision toimprove the harmonic modeling and eliminate various artifacts of the GAN-basedvocoder. Experiments show that DSPGAN significantly outperforms the comparedapproaches and it can generate high-fidelity speech for various TTS modelstrained using diverse data.

DSPGAN: a GAN-based universal vocoder for high-fidelity TTS by time-frequency domain supervision from DSP