Self-Supervised Test-Time Tuning for Packet Loss Concealment

2026-07-02

AI Summary

This work addresses the limitation of conventional packet loss concealment (PLC) models, whose fixed parameters struggle to adapt to dynamic audio signals. The authors propose TTT-PLC, a test-time training framework that fine-tunes existing PLC models online using only the received distorted audio packets, without requiring clean references, external data, or architectural modifications. TTT-PLC is the first method to enable test-time adaptation for PLC without any additional information, leveraging synthetically masked segments to generate self-supervised learning signals. It supports both causal and non-causal deployment and demonstrates consistent improvements across mainstream backbones such as FRN and PARCnet. Experimental results show that TTT-PLC significantly enhances reconstruction quality in both speech and networked music scenarios, confirming that the received signal alone can effectively drive dynamic model optimization.

Abstract

Packet loss concealment (PLC) reconstructs audio packets that are missing at the receiver, usually with a trained model whose parameters remain fixed at deployment time. This treats the PLC model as static, even though each call or recording exposes signal-specific information through the packets that did arrive. We present TTT-PLC, a self-supervised test-time tuning framework that adapts existing PLC models using only those received packets. The method creates supervision by synthetically masking portions of the available signal, training the model to conceal them with its native PLC objective, and then using the adapted model to reconstruct the true packet losses. No clean reference signal, external adaptation data, or architectural modification is required. We study TTT-PLC in two deployment settings. In the non-causal setting, the received file is available before reconstruction, allowing repeated self-supervised adaptation passes and providing a per-file adaptation ceiling. In the causal setting, audio is streamed without revising emitted samples; adaptation is performed only on completed past blocks, and updated parameters affect only future audio. We instantiate the framework on two public PLC backbones, FRN, a recurrent full-band speech PLC model, and PARCnet, a hybrid autoregressive-neural model for networked music. Across these settings, the results show that pretrained PLC systems do not need to be treated as fixed at inference time, the still-observed portions of a lossy signal can provide an effective training signal for improving concealment on that same signal.

Problem

• Packet Loss Concealment
• Self-Supervised Learning
• Test-Time Tuning
• Audio Reconstruction
• Adaptive Inference

Innovation

• test-time tuning
• self-supervised learning
• packet loss concealment
• audio reconstruction
• model adaptation