Reducing over-smoothness in speech synthesis using Generative Adversarial Networks

Leyuan Sheng, Evgeniy N. Pavlovskiy

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Abstract

Speech synthesis is widely used in many practical applications. In recent years, speech synthesis technology has developed rapidly. However, one of the reasons why synthetic speech is unnatural is that it often has over-smoothness. In order to improve the naturalness of synthetic speech, we first extract the mel-spectrogram of speech and convert it into a real image, then take the over-smooth mel-spectrogram image as input, and use image-To-image translation Generative Adversarial Networks(GANs) framework to generate a more realistic mel-spectrogram. Finally, the results show that this method greatly reduces the over-smoothness of synthesized speech and is more close to the mel-spectrogram of real speech.

Original languageEnglish
Title of host publicationSIBIRCON 2019 - International Multi-Conference on Engineering, Computer and Information Sciences, Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages972-974
Number of pages3
ISBN (Electronic)9781728144016
ISBN (Print)978-1-7281-4402-3
DOIs
Publication statusPublished - Oct 2019
Event2019 International Multi-Conference on Engineering, Computer and Information Sciences, SIBIRCON 2019 - Novosibirsk, Russian Federation
Duration: 21 Oct 201927 Oct 2019

Publication series

NameSIBIRCON 2019 - International Multi-Conference on Engineering, Computer and Information Sciences, Proceedings

Conference

Conference2019 International Multi-Conference on Engineering, Computer and Information Sciences, SIBIRCON 2019
CountryRussian Federation
CityNovosibirsk
Period21.10.201927.10.2019

Keywords

  • GANs
  • mel-spectrogram
  • over-smoothness
  • Speech synthesis

OECD FOS+WOS

  • 1.01 MATHEMATICS
  • 1.02 COMPUTER AND INFORMATION SCIENCES
  • 1.03 PHYSICAL SCIENCES AND ASTRONOMY

Fingerprint

Dive into the research topics of 'Reducing over-smoothness in speech synthesis using Generative Adversarial Networks'. Together they form a unique fingerprint.

Cite this