Speech Enhancement for Virtual Meetings on Cellular Networks

TL;DR

This paper investigates deep learning-based speech enhancement for virtual meetings over cellular networks, addressing background noise and transmission loss by creating a new dataset and evaluating two baseline models.

Contribution

It introduces a transmitted DNS dataset collected over cellular networks and compares Demucs and FullSubNet models for speech enhancement in this context.

Findings

Demucs and FullSubNet improve speech quality over cellular networks

The dataset enables realistic evaluation of speech enhancement models

Deep learning models can mitigate transmission-related speech degradation

Abstract

We study speech enhancement using deep learning (DL) for virtual meetings on cellular devices, where transmitted speech has background noise and transmission loss that affects speech quality. Since the Deep Noise Suppression (DNS) Challenge dataset does not contain practical disturbance, we collect a transmitted DNS (t-DNS) dataset using Zoom Meetings over T-Mobile network. We select two baseline models: Demucs and FullSubNet. The Demucs is an end-to-end model that takes time-domain inputs and outputs time-domain denoised speech, and the FullSubNet takes time-frequency-domain inputs and outputs the energy ratio of the target speech in the inputs. The goal of this project is to enhance the speech transmitted over the cellular networks using deep learning models.