Poisson Robust Soliton Distribution for LT Codes

Abstract

This letter proposes a novel Poisson robust soliton distribution (PRSD) for Luby transform (LT) codes. By combining the characteristics of Poisson distribution (PD) and robust soliton distribution (RSD), PRSD can improve both the decoding efficiency and the symbol recovery probability. There are three important parameters λ, δ, and c in PRSD, which directly affect the encoding/decoding performance of LT codes. The appropriate value of λ is obtained according to the mathematical properties of PD and the average degree of the coded symbols. Furthermore, based on the properties of expected ripple size, the artificial fish swarm algorithm is employed to search the appropriate values of δ and c for different-length input symbols correspondingly. Simulation results show that compared with RSD and other optimized degree distributions, the proposed degree distribution reduces overhead by at least 5.78% (5.78%-12.90%), and speeds up the encoding/decoding process by 13.4%-53.78%.