Characterizing and Modeling Multiparty Voice Communication for Multiplayer Games

Over the last few years, the number of game players using voice communication to talk to each other while playing games has increased dramatically. Unlike traditional voice-over-IP technology, where most conversations are between two people, voice communication in games often has 5 or more people talking together as they play. We present the rst measurement study on the characteristics of multiparty voice communications and develop a model of the observed talking and silence periods that can be used for future research, simulation, network engineering, and game development. Over a 3 month period, we measured over 7,000 sessions on an active multi-party voice communication server to quantify the characteristics of communication generated by game players, including group sizes, packet distributions, user and session frequencies, and speaking (and silence) durations. Unlike prior results, our measurements and models demonstrate that the speaking and silence periods t a Weibull distribution.