Unequal Error Protection Random Linear Coding Strategies for Erasure Channels

In this paper, we provide the performance analysis of unequal error protection (UEP) random linear coding (RLC) strategies designed for transmission of source messages containing packets of different importance over lossy packet erasure links. By introducing the probabilistic encoding framework, we first derive the general performance limits for the packet-level UEP coding strategies that encode the packets of each importance class of the source message independently (non-overlapping windowing strategy) or jointly (expanding windowing strategy). Then, we demonstrate that the general performance limits of both strategies are achievable by the probabilistic encoding over non-overlapping and expanding windows based on RLC and the Gaussian Elimination (GE) decoding. Throughout the paper, we present a number of examples that investigate the performance and optimization of code design parameters of the expanding window RLC strategy and compare it with the non-overlapping RLC strategy selected as a reference.