Delay jitter compensation for low latecy audio in Networked Music Performances

Abstract

In this thesis an algorithm to compensate for delay jitter is developed in the context of Networked Music Performances, with the goal of mitigating dropouts due to delay jitter with the minimal amount of added latency. To achieve this, the project is developed in a proprietary network simulator in which some novel tools are introduced. These tools include efficient running statistical measures such as the mean, an optimized histogram, and percentile calculation. The resulting jitter compensation algorithm is capable of determining the optimal jitter buffer delay that yields the lowest latency while adhering to a predefined threshold for acceptable dropouts. Remarkably, this algorithm is agnostic to the underlying statistical distribution of delay or the selected error ratio threshold, making it adaptable to diverse network conditions. Furthermore, we derive a canonical expression that simplifies the algorithm the optimality of which is demonstrated through empirical evaluations.