Virtual Audio Chat: User Interaction and Audio Quality Evaluation

Lea Skorin-Kapov, Saska Matesa, Ericsson Nikola Tesla

Virtual Audio Chat is an interactive Virtual Reality (VR)/Web application that enables real-time audio streaming between multiple Internet users. Users join the audio chat using a VR interface designed as a Virtual Reality Modeling Language (VRML) model of a mobile phone. Various interactions implemented using VRML allow users to choose the color/textures applied to the mobile phone and the audio codec used during the audio streaming session.

In this paper we address the developed application in terms of user interactions with the VR interface and evaluation of audio quality. Our primary focus is to evaluate and compare audio quality for different audio codecs. Overall speech quality and network requirements depend in great deal on the type of codec being implemented. Different codecs result in different data rates and Mean Opinion Scores (MOS) relating to subjective quality. We performed measurements of network throughput and subjective speech quality for three different codecs: PCM, GSM, and G.723.1. Two sets of measurements were performed: the first set using previously recorded speech and music, and the second set during a live chat between two participants. Results have been analyzed and the codecs have been compared in terms of network throughput and achieved subjective speech quality.