APSIPA Transactions on Signal and Information Processing > Vol 11 > Issue 1

A Cyclical Approach to Synthetic and Natural Speech Mismatch Refinement of Neural Post-filter for Low-cost Text-to-speech System

Yi-Chiao Wu, Information Technology Center, Nagoya University, Japan, yichiao.wu@g.sp.m.is.nagoya-u.ac.jp , Patrick Lumban Tobing, Information Technology Center, Nagoya University, Japan, Kazuki Yasuhara, AI, Inc., Japan, Noriyuki Matsunaga, AI, Inc., Japan, Yamato Ohtani, AI, Inc., Japan, Tomoki Toda, Information Technology Center, Nagoya University, Japan
 
Suggested Citation
Yi-Chiao Wu, Patrick Lumban Tobing, Kazuki Yasuhara, Noriyuki Matsunaga, Yamato Ohtani and Tomoki Toda (2022), "A Cyclical Approach to Synthetic and Natural Speech Mismatch Refinement of Neural Post-filter for Low-cost Text-to-speech System", APSIPA Transactions on Signal and Information Processing: Vol. 11: No. 1, e30. http://dx.doi.org/10.1561/116.00000020

Publication Date: 21 Sep 2022
© 2022 Y.-C. Wu, P. L. Tobing, K. Yasuhara, N. Matsunaga, Y. Ohtani and T. Toda
 
Subjects
 
Keywords
Neural post-filter for text-to-speechtemporal mismatchacoustic mismatchneural vocodercyclical voice conversion
 

Share

Open Access

This is published under the terms of CC BY-NC.

Downloaded: 127 times

In this article:
Introduction 
Related Work 
Neural Vocoder 
Cyclical Spectral Conversion 
Proposed Cyclical Approach to Developing Neural Post-filter for TTS 
Experiment with Oracle Phoneme Duration 
Experiment with Predicted Phoneme Duration 
Conclusion 
References 

Abstract

Neural-based text-to-speech (TTS) systems achieve very high-fidelity speech generation because of the rapid neural network developments. However, the huge labeled corpus and high computation cost requirements limit the possibility of developing a high-fidelity TTS system by small companies or individuals. On the other hand, a neural vocoder, which has been widely adopted for the speech generation in neural-based TTS systems, can be trained with a relatively small unlabeled corpus. Therefore, in this paper, we explore a general framework to develop a neural post-filter (NPF) for low-cost TTS systems using neural vocoders. A cyclical approach is proposed to tackle the acoustic and temporal mismatches (AM and TM) of developing an NPF. Both objective and subjective evaluations have been conducted to demonstrate the AM and TM problems and the effectiveness of the proposed framework.

DOI:10.1561/116.00000020