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 1909.07352eess.AScs.SDeess.SP
收藏

基于两级多模态网络的音视频语音分离与去混响

Audio-Visual Speech Separation and Dereverberation with a Two-Stage Multimodal Network

The Ohio State University(俄亥俄州立大学);Tencent AI Lab(腾讯AI实验室)

Ke Tan, Yong Xu, Shi-Xiong Zhang, Meng Yu, Dong Yu

摘要:

在真实的听音环境中,背景噪声、干扰语音和室内混响常常会使目标语音失真。在这项研究中,我们解决联合语音分离和去混响,其目的是从背景噪声,干扰语音和室内混响分离目标语音。为了解决这个根本性的难题,我们提出了一种新的多模态网络,利用音频和视觉信号。该网络结构采用两阶段策略,第一阶段采用分离模块衰减背景噪声和干扰语音,第二阶段采用去混响模块抑制室内混响。这两个模块首先单独训练,然后集成的联合训练,这是基于一个新的多目标损失函数。我们的实验结果表明,所提出的多模态网络产生一贯更好的客观可懂度和感知质量比几个一阶段和两阶段的基线。我们发现,我们的网络实现了21.10%的改善,在ESTOI和0.79改善PESQ比未经处理的混合物。此外,我们的网络架构不需要知道扬声器的数量。

英文摘要:

Background noise, interfering speech and room reverberation frequently distort target speech in real listening environments. In this study, we address joint speech separation and dereverberation, which aims to separate target speech from background noise, interfering speech and room reverberation. In order to tackle this fundamentally difficult problem, we propose a novel multimodal network that exploits both audio and visual signals. The proposed network architecture adopts a two-stage strategy, where a separation module is employed to attenuate background noise and interfering speech in the first stage and a dereverberation module to suppress room reverberation in the second stage. The two modules are first trained separately, and then integrated for joint training, which is based on a new multi-objective loss function. Our experimental results show that the proposed multimodal network yields consistently better objective intelligibility and perceptual quality than several one-stage and two-stage baselines. We find that our network achieves a 21.10% improvement in ESTOI and a 0.79 improvement in PESQ over the unprocessed mixtures. Moreover, our network architecture does not require the knowledge of the number of speakers.

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。

↑