PLC Testbench: a modular tool for the study and comparison of audio Packet Loss Concealment algorithms

Whenever time-critical applications supporting re-mote interactions are deployed on the Internet they may face the issue of packet loss. This is particularly true for remote interactions that are audio-based and require low latency and high reliability, such as those enabled by networked music performance systems, where packet losses are detrimental to the quality of experience of the users. Packet Loss Concealment (PLC) for audio signals is then mandatory to cope with such an issue. PLC is a technique that aims at masking the effect of packet loss on human auditory perception. Many PLC algorithms have been developed over the last three decades, initially focusing on speech and more recently addressing broadband audio. However, to date, developers of PLC methods face the issue of the lack of tools supporting the design, comparison, and testing of such methods. To bridge this gap, this paper introduces PLC Testbench, a software tool that enables researchers to study, compare, and evaluate various PLC algorithms for audio. The software is modular and comprises a Python backend and a web user interface. It is open-source and is conveniently distributed as a docker container.