Progressive Parallel WaveGAN 


Vol. 46,  No. 10, pp. 1596-1603, Oct.  2021
10.7840/kics.2021.46.10.1596


PDF
  Abstract

In this paper, we propose a progressive parallel waveNet and apply it as a generator to parallel waveGAN, a speech synthesis vocoder, to confirm the performance of the proposed model. Parallel waveNet, a generator of parallel waveGAN showing high performance in speech synthesis vocoder, requires a lot of GPUs for training. To solve this problem, the parallel wavenet structure was transformed into a progressive structure using an up-sampling process of speech. This is a structure that effectively solves the large dimension problem of the parallel wavenet. In addition, it is a structure that can increase the stability and performance of learning by gradually estimating a wide frequency band of speech. As a result of the experiment, it was confirmed that less GPU usage on training and faster inference speed compared to parallel waveGAN, and inference speed was also faster. Finally, when speech quality was measured as an objective measurement, a slightly improved result was also confirmed.

  Statistics
Cumulative Counts from November, 2022
Multiple requests among the same browser session are counted as one view. If you mouse over a chart, the values of data points will be shown.


  Cite this article

[IEEE Style]

H. Y. Kim, B. J. Woo, N. S. Kim, "Progressive Parallel WaveGAN," The Journal of Korean Institute of Communications and Information Sciences, vol. 46, no. 10, pp. 1596-1603, 2021. DOI: 10.7840/kics.2021.46.10.1596.

[ACM Style]

Hyung Yong Kim, Beom Jun Woo, and Nam Soo Kim. 2021. Progressive Parallel WaveGAN. The Journal of Korean Institute of Communications and Information Sciences, 46, 10, (2021), 1596-1603. DOI: 10.7840/kics.2021.46.10.1596.

[KICS Style]

Hyung Yong Kim, Beom Jun Woo, Nam Soo Kim, "Progressive Parallel WaveGAN," The Journal of Korean Institute of Communications and Information Sciences, vol. 46, no. 10, pp. 1596-1603, 10. 2021. (https://doi.org/10.7840/kics.2021.46.10.1596)

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)






pISSN : 1226 - 4717
eISSN : 2287 - 3880