Characterization of Luby Transform Codes with Small Message Size for Low-Latency Decoding

Luby transform (LT) codes provide efficient erasure protection for information distributed across a multiuser channel with unknown characteristics. For real-time applications, however, the random selection of information symbols within a large message can increase encoding and decoding latency beyond allowable parameters. By using a smaller message size, the encoder can generate a codeword at a faster rate, leading to a higher decoder throughput. We decrease decoding latency by optimizing various parameters of the Robust Soliton distribution. This optimization in terms of delay does not lead to minimizing overhead. However, if a small increase in overhead can be tolerated, particularly for a smaller message size, then the overall encoding and decoding delay can be greatly reduced.