Characterization of Arrival Statistics of Multiplexed Voice Packets

An assumption that voice packets arrive by a Poisson arrival (or an exponential interarrival) distribution has not widely been accepted by analysts from an obvious observation that voice packets are generated at regular intervals in a talkspurt. Through a simulation, supported by an approximate analysis, this paper claims that when a sufficiently large number of voice sources are multiplexed, packets observe an exponential or hypoexponential interarrival distribution. When the number of packets arriving during a single or a multiple slot is considered instead, a more liberal claim is made that regardless of the number of voice sources multiplexed, a Poisson process can be assumed with reasonable accuracy.