An approach to the quantification of delay variation effects on live audio streams

The aim of the paper is to propose an approach to studying delay variation effects on live voice streams and mapping them to quantifiable parameters. A suitable performance evaluation framework is defined within which an appropriate metric, based on the (m,k) firm guarantee criterion, is studied and tailored to interactive voice streams. This metric is based on the observation that different competing traffic patterns, resulting in the same delay variation probability distribution functions of the voice stream, may have a different impact on the perceived quality, thus corresponding to different levels of quality. The capability of the proposed metric to capture delay variation sensitivities of the stream is illustrated through experiments. Then, the performance evaluation framework is applied through scenarios where the voice stream shares the same link along with bursty competing traffic composed of several multiplexed VBR and CBR video streams. The network is appropriately configured so that only delay variation and no losses are experienced in the queuing process. Experimental measurements are disclosed and conclusions are derived concerning the suitability of the proposed metric for evaluating the coexistence of live voice streams with bursty competing traffic.