Comparison of Conventional and Cross-Layer Multimedia Transport Schemes for Wireless Networks

Three competing schemes have been proposed for multimedia transport over broadband wireless channels: (a) traditional UDP (Postel, The User Datagram Protocol, 1980 [1]), (b) semi-cross-layer UDP-Lite (The Lightweight User Datagram Protocol, 2004 [2]), and (c) cross-layer header estimation (Khayam et al., IEEE Transactions on Multimedia 9(2):377–385, 2007 [3]; Khayam and Radha, IEEE Transactions on Wireless Communications 6(11):3946–3954, 2007 [4]). In all these schemes, corrupted and lost packets are recovered using FEC at the application layer. In this paper, we analytically and experimentally compare the performances of these broadband wireless multimedia schemes. First, we derive lower bounds on the excepted FEC redundancy required by ideal cross-layer header estimation, UDP and UDP-Lite over an arbitrary-order Markov wireless channel. We show that under realistic wireless channel conditions, the cross-layer header estimation scheme always requires lesser redundancy than UDP and UDP-Lite. We then propose a practical minimum distance decoding (MDD) header estimation scheme, which is receiver-based, low complexity and highly accurate. Trace-driven multimedia experiments over wireless LANs demonstrate that MDD header estimation requires significantly lesser FEC redundancy and renders better video quality than existing schemes.