Duration: 38:30

170 views

2

We introduce a new structure for multi-party audio conferencing, where the MCU (Multipoint Conferencing Unit) mixes all the received signals and rebroadcast the same stream to all clients. Each client then removes its own signal before playback. This architecture reduces bandwidth and computation requirements at the MCU. The major challenge, however, is that the client's own speech is non-linearly distorted at the MCU. This makes it hard to remove from the mixed signal, producing an echo-like distortion. In this work, we solve that problem by proposing a client-side echo cancellation algorithm which completely removes the echo, with minimal artifacts. The Mean Opinion Score (MOS) results imply that the proposed algorithm works well, which makes the proposed multi-party audio conferencing structure practical and promising.