Performance modeling of variable bit rate voice in integrated packet networks

The authors present methodologies for evaluating the performance of variable-bit-rate voice under the following conditions: at a fixed load when instantaneous fluctuations occur due to talker activity/inactivity, and (2) under variable load when variations occur due to call on/off. A Markov chain model is developed to describe the probabilistic bit-dropping pattern corresponding to any specific traffic conditions. The model is used in conjunction with a software tool to emulate packetized voice under various loading conditions. The results of the subjective testing show that, with prudential traffic engineering of the network, the voice quality would remain robust to temporal variations in bits per sample resulting from nominal load fluctuations.<
>