|Title||Optimized network-adaptive multimedia transmission over packet erasure channels|
|Abstract||The transmission of delay-sensitive multimedia data over packet erasure channels like the Internet imposes a challenge because these networks provide transportation service in a best-effort manner only and do not guarantee the delivery of the packets in time. Packets which are lost or arrive at the receiver after their presentation deadline have devastating effects on the perceived quality of the multimedia. This problem can be circumvented with forward error control (FEC), automatic repeat request (ARQ), error resilient coding, error concealment, or any combination thereof.
While TCP (Transmission Control Protocol) works well for non real-time data, e.g., ﬁle transfer, its relevance for real-time or delay-sensitive applications is debatable. This is the case, for example, when a user wants to download a three dimensional (3D) mesh in a certain time period or in video streaming where each video frame has to be delivered and displayed within a certain time period to enable continuous playback. For such applications, FEC may be a better candidate.
The ﬁrst part of this dissertation considers application-layer FEC to protect progressively compressed 3D meshes against packet losses. We proposed an efficient algorithm that, for a given total transmission bit budget, ﬁnds the optimal source-channel bit allocation in linear time and space. The proposed algorithm can be used to protect, in general, any progressively compressed layered source data.
The second part of this dissertation presents a video streaming framework that uses application-layer FEC to mitigate the eﬀect of lost packets. In previous work, the FEC code-rate is either ﬁxed or updated periodically according to an estimated packet loss rate. However, this approach may not perform well when the packet loss rate ﬂuctuates rapidly over time and is, therefore, hard to estimate with enough accuracy. Indeed, underestimating the packet loss rate leads to code failure and overestimating it would misspend the bandwidth. To deal with the ﬂuctuating packet loss rates, we introduce a novel approach based on rateless codes and receiver feedback. Within this framework, we present efficient transmission strategies and an efficient algorithm to minimize the bandwidth usage while ensuring successful decoding subject to an upper bound on the packet loss rate. Analytical and simulated results show that, compared to traditional FEC based approaches, the proposed scheme provides signiﬁcant bandwidth savings for the same playback quality. In an experiment, where H.264 encoded Foreman sequence was transmit- ted over an Internet channel (Konstanz-Beijing-Konstanz), our proposed solution outperformed the other FEC based approaches e.g., the PSNR gain over ﬁxed-rate FEC scheme exceeds 3.5 decibels at 90 kbps.