arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2302.14036eess.AScs.CLcs.LGcs.SD
收藏

使用集成文本到梅尔频谱生成器的端到端ASR纯文本域自适应

Text-only domain adaptation for end-to-end ASR using integrated text-to-mel-spectrogram generator

NVIDIA(英伟达);University of London(伦敦大学);ITMO University(ITMO大学);Higher School of Economics(高等经济学院)

Vladimir Bataev, Roman Korostik, Evgeny Shabalin, Vitaly Lavrukhin, Boris Ginsburg

AI总结:

本文提出一种端到端ASR系统,通过集成非自回归多说话人文本到梅尔频谱生成器和GAN增强器,利用纯文本数据实现领域自适应,在准确率、自适应质量和训练速度上均优于仅用转写语音训练及级联TTS方案。

摘要:

我们提出了一种端到端自动语音识别(ASR)系统,该系统可以在有转写的语音数据、纯文本数据或两者的混合数据上进行训练。所提出的模型使用一个集成的辅助模块进行基于文本的训练。该模块将非自回归多说话人文本到梅尔频谱生成器与基于GAN的增强器相结合,以提高频谱质量。所提出的系统可以在训练过程中动态生成梅尔频谱。它可以通过使用来自新领域的纯文本数据,将ASR模型自适应到该新领域。我们证明,与仅在转写语音上训练的系统相比,所提出的训练方法显著提高了ASR准确率。在自适应质量和训练速度方面,它也超过了带声码器的级联TTS系统。

英文摘要:

We propose an end-to-end Automatic Speech Recognition (ASR) system that can be trained on transcribed speech data, text-only data, or a mixture of both. The proposed model uses an integrated auxiliary block for text-based training. This block combines a non-autoregressive multi-speaker text-to-mel-spectrogram generator with a GAN-based enhancer to improve the spectrogram quality. The proposed system can generate a mel-spectrogram dynamically during training. It can be used to adapt the ASR model to a new domain by using text-only data from this domain. We demonstrate that the proposed training method significantly improves ASR accuracy compared to the system trained on transcribed speech only. It also surpasses cascade TTS systems with the vocoder in the adaptation quality and training speed.

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。

↑