Year: 2022
Type: conference-paper
Abstract: Neural network based end-to-end Text-to-Speech (TTS) has greatly improved the quality of synthesized speech. However, (the efficient usage of massive)[how to efficiently use massive] spontaneous speech without transcription remains an open problem. In this paper, we propose MHTTS, a fast multi-speaker TTS system that is robust to transcription errors of real speech data. Specifically, we introduce a multi-head model and a method that transfer text information from corpora with manual transcription to corpora without annotation. MHTTS has two advantages: 1) Our system synthesizes better quality multi-speaker voice with faster inference speed. 2) Our system is capable of transferring correct text information to data with imperfect transcription, simulated using corruption, or provided by an Automatic Speech Recognizer (ASR).
Authors: Dabiao Ma, Zhiba Su, Yitong Zhang, Emily Huang, Meng Li, Qiwen Lyu, Feng Ye
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4366308362