A Study of Silence Suppression and Real Speech Patterns and their Impact on VoIP Capacity in 802.11 Networks

We present a study of voice over IP capacity in 802.11 wireless local area networks (WLANs) using actual conversational speech recordings. Our intention is to give a more realistic view on how voice capacity varies in a WLAN environment. To this goal we conducted several parallel studies comparing results based on real speech to those based on the ITU-T Rec. P.59 artificial conversational model. The studies include the use (or not) of silence suppression, distributed and centralized access mechanisms, and a variety of speech conditions. The results show, importantly, that indeed capacity based on real speech differs from that obtained using the artificial model. The reasons for this difference goes beyond observed differences in activity levels. It depends on secondary statistical characteristics of traffic and on the (random) sampling of conversations. With the study we are also able to compare the relative benefits of different approaches to increase capacity.