Error recovery for multicast conversational video over error-prone networks (with application to ad hoc WLANs)

Packet error rates in multicast conversational video transmissions over error-prone networks severely affect the performance of motion-compensated hybrid decoders. The loss of information in one frame has a considerable impact on the quality of the following frames, and errors propagate in the sequence. Conversational video requires fast error recovery, but multicast transmission with real-time interactive constraints introduces serious difficulties. We introduce a feedback-based protocol for resilient multicast videoconferencing in error-prone LANs. In particular, we focus on video transmission over ad hoc wireless LANs (WLANs). Unlike previous approaches, our scheme not only improves video quality at error-prone receivers, but also maximizes the overall video quality of the session and achieves efficient use of the available bandwidth. This protocol uses a decentralized repair request algorithm with (i) a new multicast suppression technique that protects critical correction requests and eliminates duplicated or overlapped control packets, and (ii) a new error resilience scheme based on multicast repair packets that convey predictive correction information to the video session. We present experimental videoconferencing results that demonstrate the effectiveness of the proposed protocol in ad hoc WLANs.