Reducing over-smoothness in speech synthesis using Generative Adversarial Networks

Sheng, Leyuan; Pavlovskiy, Evgeniy

doi:10.48550/arxiv.1810.10989

Search citation statements

Order By: Relevance

Paper Sections

Select...

Introduction1

Citation Types

Supporting

Mentioning

Contrasting

Year Published

2019

Publication Types

Select...

Other1

Relationship

Self Cite1

Independent0

Authors

Journals

Cited by 1 publication

(1 citation statement)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, the conditional GAN (cGAN) architecture adopted a pixel-to-pixel (Pix2Pix) framework has investigated for SE [12] and Music Information Retrieval (MIR) [13]. As the high-resolution image synthesis architecture Pix2PixHD [14] improves the performance than Pix2Pix, in [15] the authors have implemented the Pix2PixHD within a cGAN framework for reducing oversmoothness in speech synthesis.…”

Section: Introductionmentioning

confidence: 99%

High-quality Speech Synthesis Using Super-resolution Mel-Spectrogram

Sheng¹,

Huang²,

Pavlovskiy³

2019

Preprint

Self Cite

View full text Add to dashboard Cite

In speech synthesis and speech enhancement systems, melspectrograms need to be precise in acoustic representations. However, the generated spectrograms are over-smooth, that could not produce high quality synthesized speech. Inspired by image-to-image translation, we address this problem by using a learning-based post filter combining Pix2PixHD and ResUnet to reconstruct the mel-spectrograms together with super-resolution. From the resulting super-resolution spectrogram networks, we can generate enhanced spectrograms to produce high quality synthesized speech. Our proposed model achieves improved mean opinion scores (MOS) of 3.71 and 4.01 over baseline results of 3.29 and 3.84, while using vocoder Griffin-Lim and WaveNet, respectively.

show abstract