CPRSD for LT codes

Abstract

Degree distribution has a great influence on the performance of Luby Transform (LT) codes. In this study, the authors study a new degree distribution for LT codes. First, improved Poisson distribution (IPD) is proposed by refining the PD. The successful decoding rate for IPD reaches more than 90% when the decoder receives less encoding symbols, while robust soliton distribution (RSD) performs better as received encoding symbols grow. Then they combine IPD and RSD by a parameter to derive the function of combined Poisson RSD (CPRSD). Moreover, the golden section algorithm is introduced to obtain the optimal parameter for CPRSD. The expected ripple size is employed to present the objective function for this algorithm, and their objective is to increase the mean of ripple size and reduce its variation at the same time. Simulation results show that when compared with RSD and other optimised degree distributions, the CPRSD achieves at least 20.1% (20.1%–30.4%) reduction in overhead, and saves 28%–46.2% consuming time per encoding/decoding process.