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

This paper presents a low-latency real-time (LLRT) non-parallel voiceconversion (VC) framework based on cyclic variational autoencoder (CycleVAE)and multiband WaveRNN with data-driven linear prediction (MWDLP). CycleVAE is arobust non-parallel multispeaker spectral model, which utilizes aspeaker-independent latent space and a speaker-dependent code to generatereconstructed/converted spectral features given the spectral features of aninput speaker. On the other hand, MWDLP is an efficient and a high-qualityneural vocoder that can handle multispeaker data and generate speech waveformfor LLRT applications with CPU. To accommodate LLRT constraint with CPU, wepropose a novel CycleVAE framework that utilizes mel-spectrogram as spectralfeatures and is built with a sparse network architecture. Further, to improvethe modeling performance, we also propose a novel fine-tuning procedure thatrefines the frame-rate CycleVAE network by utilizing the waveform loss from theMWDLP network. The experimental results demonstrate that the proposed frameworkachieves high-performance VC, while allowing for LLRT usage with a single-coreof $2.1$--$2.7$ GHz CPU on a real-time factor of $0.87$--$0.95$, includinginput/output, feature extraction, on a frame shift of $10$ ms, a window lengthof $27.5$ ms, and $2$ lookup frames.

Low-Latency Real-Time Non-Parallel Voice Conversion based on Cyclic Variational Autoencoder and Multiband WaveRNN with Data-Driven Linear Prediction