arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1905.09263cs.CLcs.LGcs.SDeess.AS
收藏

FastSpeech:快速、健壮、可控的文本到语音转换

FastSpeech: Fast, Robust and Controllable Text to Speech

Zhejiang University(浙江大学);Microsoft Research(微软研究院);Microsoft STC Asia(微软亚洲互联网工程院)

Yi Ren, Yangjun Ruan, Xu Tan, Tao Qin, Sheng Zhao, Zhou Zhao, Tie-Yan Liu

摘要:

基于神经网络的端到端文语转换(TTS)技术显著提高了合成语音的质量。突出的方法(例如,Tacotron 2)通常首先从文本生成梅尔频谱图,然后使用声码器(如WaveNet)从梅尔频谱图合成语音。与传统的拼接和统计参数方法相比,基于神经网络的端到端模型的推理速度慢,并且合成的语音通常不鲁棒(即,某些单词被跳过或重复)和缺乏可控性(语音速度或韵律控制)。在这项工作中,我们提出了一种新的前馈网络的基础上Transformer的并行生成梅尔频谱的TTS。具体而言,我们提取注意力对齐从编码器-解码器为基础的教师模型音素持续时间预测,这是由长度调节器扩展源音素序列,以匹配的目标梅尔频谱图序列的长度并行梅尔频谱图生成。在LJSpeech数据集上的实验表明,我们的并行模型在语音质量方面与自回归模型相匹配,几乎消除了特别困难情况下的单词跳过和重复问题,并且可以平滑地调整语音速度。最重要的是,与自回归Transformer TTS相比,我们的模型将梅尔频谱图生成速度提高了270倍,端到端语音合成速度提高了38倍。因此,我们称我们的模型为FastSpeech。

英文摘要:

Neural network based end-to-end text to speech (TTS) has significantly improved the quality of synthesized speech. Prominent methods (e.g., Tacotron 2) usually first generate mel-spectrogram from text, and then synthesize speech from the mel-spectrogram using vocoder such as WaveNet. Compared with traditional concatenative and statistical parametric approaches, neural network based end-to-end models suffer from slow inference speed, and the synthesized speech is usually not robust (i.e., some words are skipped or repeated) and lack of controllability (voice speed or prosody control). In this work, we propose a novel feed-forward network based on Transformer to generate mel-spectrogram in parallel for TTS. Specifically, we extract attention alignments from an encoder-decoder based teacher model for phoneme duration prediction, which is used by a length regulator to expand the source phoneme sequence to match the length of the target mel-spectrogram sequence for parallel mel-spectrogram generation. Experiments on the LJSpeech dataset show that our parallel model matches autoregressive models in terms of speech quality, nearly eliminates the problem of word skipping and repeating in particularly hard cases, and can adjust voice speed smoothly. Most importantly, compared with autoregressive Transformer TTS, our model speeds up mel-spectrogram generation by 270x and the end-to-end speech synthesis by 38x. Therefore, we call our model FastSpeech.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑