•  
  •  
 

Abstract

This paper presents a novel packet loss concealment system called “Regulator,” designed for interactive network audio applications, particularly server-mediated “jam rooms” in the cloud, where large ensembles of musicians perform together synchronously. The system addresses the critical challenge of maintaining continuous high-quality audio output while minimizing latency penalties in unreliable network environments including Wi-Fi. The Regulator architecture combines four core components: a central Regulator class orchestrating packet loss concealment operations; a BurgAlgorithm class implementing maximum entropy autoregressive prediction to reconstruct missing audio samples; a Channel class providing per-channel independent modeling for multi-channel streams; and a RegulatorWorker class offering asynchronous processing through lock-free threading. The system’s packet selection logic evaluates multiple candidate packets to determine optimal playback timing, and its cross-fading mechanisms ensure smooth transitions between predicted and real audio to minimize audible artifacts. An adaptive tolerance management system automatically adjusts buffer headroom based on observed network jitter patterns. Unlike traditional approaches that add significant delay or produce glitchy substitutes, Regulator passes received streams through with minimal added delay while providing high-quality packet loss concealment. It supports various audio formats—including different sample rates, bit depths, frame sizes, and multi-channel configurations—enabling deployment across heterogeneous network audio scenarios.

License Information

This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License (CC BY-NC-ND 4.0).

Download

Share

COinS