Speech Quality Evaluation for Large-Scale D2D (Device to Device) VoIP Conference

Abstract

Several models and architectures are proposed to support multiparty VoIP (MVoIP) service. Even if some of these solutions support large number of participants using a centralized conference server, solutions that base their media distribution on peers are adapted to support only small group of conferences. We introduced a conference architecture framework controlled and administrated in centralized manner while media distribution is fully distributed on participating peers within a tree-based model. A realistic VoIP simulator based on Omnet environment has been introduced to this work. Thus, several media coding/decoding on media tree levels are processed using different coding rates. The quality of the received speech and consumed bandwidth are assessed and compared depending on the traversed media tree death of the conference participating members and the silence threshold.