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

GAN vocoders are currently one of the state-of-the-art methods for buildinghigh-quality neural waveform generative models. However, most of theirarchitectures require dozens of billion floating-point operations per second(GFLOPS) to generate speech waveforms in samplewise manner. This makes GANvocoders still challenging to run on normal CPUs without accelerators orparallel computers. In this work, we propose a new architecture for GANvocoders that mainly depends on recurrent and fully-connected networks todirectly generate the time domain signal in framewise manner. This results inconsiderable reduction of the computational cost and enables very fastgeneration on both GPUs and low-complexity CPUs. Experimental results show thatour Framewise WaveGAN vocoder achieves significantly higher quality thanauto-regressive maximum-likelihood vocoders such as LPCNet at a very lowcomplexity of 1.2 GFLOPS. This makes GAN vocoders more practical on edge andlow-power devices.

Framewise WaveGAN: High Speed Adversarial Vocoder in Time Domain with Very Low Computational Complexity