Improving performance of real-time full-band blind packet-loss concealment with predictive network

TL;DR

This paper introduces a real-time recurrent neural network for full-band packet-loss concealment in high-quality telecommunication, significantly improving speech quality during network disruptions.

Contribution

It presents a novel full-band recurrent network that operates at 48 kHz for real-time PLC without needing prior loss information, outperforming existing methods.

Findings

FRN outperforms offline non-causal baselines

FRN ranks top in recent PLC challenge

Real-time operation at 48 kHz enhances telecommunication quality

Abstract

Packet loss concealment (PLC) is a tool for enhancing speech degradation caused by poor network conditions or underflow/overflow in audio processing pipelines. We propose a real-time recurrent method that leverages previous outputs to mitigate artefact of lost packets without the prior knowledge of loss mask. The proposed full-band recurrent network (FRN) model operates at 48 kHz, which is suitable for high-quality telecommunication applications. Experiment results highlight the superiority of FRN over an offline non-causal baseline and a top performer in a recent PLC challenge.