arXiv：cs.LG· Wanli Sun, Zehai Tu, Anton Ragni·· 4 小时前
基于能量模型的语音合成
Energy-Based Models For Speech Synthesis
AI 导读
论文将能量模型（EBM）引入非自回归语音合成，提出用噪声对比估计训练 EBM，并给出多种生成有效负样本的策略（包括借助高性能 AR 模型），采样则采用 Langevin MCMC，从而建立 EBM 与扩散模型的联系。在 LJSpeech 数据集上，该方法相较 Tacotron 2 取得改进。
正文
Abstract:Recently there has been a lot of interest in non-autoregressive (non-AR) models for speech synthesis, such as FastSpeech 2 and diffusion models. Unlike AR models, these models do not have autoregressive dependencies among outputs which makes inference efficient. This paper expands the range of available non-AR models with another member called energy-based models (EBMs). The paper describes how noise contrastive estimation, which relies on the comparison between positive and negative samples, can be used to train EBMs. It proposes a number of strategies for generating effective negative samples, including using high-performing AR models. It also describes how sampling from EBMs can be performed using Langevin Markov Chain Monte-Carlo (MCMC). The use of Langevin MCMC enables to draw connections between EBMs and currently popular diffusion models. Experiments on LJSpeech dataset show that the proposed approach offers improvements over Tacotron 2.
| Comments: | 5 pages, 1 figure, 6 tables. Accepted to ICASSP 2024 |
| Subjects: | Sound (cs.SD); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS) |
| Cite as: | arXiv:2310.12765 [cs.SD] |
| (or arXiv:2310.12765v2 [cs.SD] for this version) | |
| https://doi.org/10.48550/arXiv.2310.12765 arXiv-issued DOI via DataCite |
Submission history
From: Wanli Sun [view email]
[v1]
Thu, 19 Oct 2023 14:10:09 UTC (86 KB)
[v2]
Wed, 14 Oct 2026 07:19:31 UTC (86 KB)
来源：arXiv：cs.LG · arxiv.org