Endpoints or conference servers of current audio-conferencing solutions use all the audio frames they receive in order to mix them into one final aggregate stream. However, at each time-instant, some of this content may not be audible due to auditory masking. Hence, sending corresponding frames through the network leads to a loss of bandwidth, while decoding them for mixing or spatial audio processing leads to increased processor load. In this paper, we propose a solution based on an efficient on-the-fly auditory masking evaluation. Our technique allows prioritizing audio frames in order to select only those audible for each connected client. We present results of quality tests showing the transparency of the algorithm. We describe its integration in a France Telecom audio conference server. Tests in a 3D game environment with spatialized chat capabilities show a 70% average reduction in required bandwidth, demonstrating the efficiency of our method.
(See document for exact affiliation information.)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
