arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2301.09027cs.SDcs.LGeess.AS
收藏

蜂窝网络语音增强:去除背景和传输噪声

Cellular Network Speech Enhancement: Removing Background and Transmission Noise

Carnegie Mellon University(卡内基梅隆大学)

Amanda Shu, Hamza Khalid, Haohui Liu, Shikhar Agnihotri, Joseph Konan, Ojas Bhargave

摘要:

语音增强的主要目标是在保留目标语音的同时降低背景噪声。当说话者被限制在嘈杂的环境中并且接收具有高背景和传输噪声的呼叫时,常见的困境发生。为了解决这个问题,深度噪声抑制(DNS)挑战赛的重点是使用下一代深度学习模型消除背景噪声,以增强目标的语音;然而,研究人员未能考虑IP语音(VoIP)应用的传输噪声。我们专注于Google Meet及其蜂窝应用程序,在VoIP DNS挑战赛的Google Meet To Phone Track上实现了最先进的性能。本文演示了如何击败工业性能,实现1.92 PESQ和0.88 STOI,以及各种指标的卓越声学保真度,感知质量和可懂度。

英文摘要:

The primary objective of speech enhancement is to reduce background noise while preserving the target's speech. A common dilemma occurs when a speaker is confined to a noisy environment and receives a call with high background and transmission noise. To address this problem, the Deep Noise Suppression (DNS) Challenge focuses on removing the background noise with the next-generation deep learning models to enhance the target's speech; however, researchers fail to consider Voice Over IP (VoIP) applications their transmission noise. Focusing on Google Meet and its cellular application, our work achieves state-of-the-art performance on the Google Meet To Phone Track of the VoIP DNS Challenge. This paper demonstrates how to beat industrial performance and achieve 1.92 PESQ and 0.88 STOI, as well as superior acoustic fidelity, perceptual quality, and intelligibility in various metrics.

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。

↑