Maximum-Likelihood Header Estimation: A Cross-Layer Methodology for Wireless Multimedia

We propose a novel cross-layer header estimation methodology that can be used by UDP-based wireless multimedia applications to estimate corrupted packet headers, thereby realizing significant throughput improvements. The proposed methodology requires only minor modifications to the protocol stack at the receiver while no modifications are needed to senders or intermediate nodes. We formulate header estimation as a problem of maximum-likelihood estimation of known parameters in noise. We derive likelihood functions for two wireless channel models, namely Markov and multifractal wavelet models. Our trace-driven video simulations at 2, 5.5 and 11 Mbps data rates of an 802.11b LAN demonstrate that significant improvements over normal UDP and UDP Lite can be achieved by employing header estimation with UDP.