Source Models for Speech Traffic Revisited

Michael Menth, Andreas Binzenhöfer, and Stefan Mühleck.
Research Report 426

Abstract

In this paper, we analyze output traces of different voice codecs and present analytical models to describe them by stochastic processes. Both the G.711 and the G.729.1 codec yield constant bit rate traffic, the G.723.1 as well as the iLBC codec use silence detection leading to an on/off process, and the GSM AMR and the iSAC codec produce variable bit rate (VBR) traffic. We apply all codecs to a large set of typical speech samples and provide quantitative models that are based on standard and modified on/off processes as well as memory Markov chains. Our models are in good accordance with the original traces as they capture the complementary cumulative distribution function (CCDF) of the on/off phase durations and the packet sizes, the autocorrelation function (ACF) of consecutive packet sizes, and the queuing properties of the original traces. In addition, they are rather simple which makes them especially useful for application in analytical and simulative studies. The durations of the on/off phases in our model are an order of magnitude larger than those found by previous work and to the best of our knowledge this is the first paper presenting a model for VBR voice.