Summary
The 2018 International Symposium on Information Theory and Its Applications　(ISITA2018)
2018
Session Number:We-AM-Poster
Session:
Number:We-AM-Poster.2
Improved LT Code Degree Distribution and its Performance Evaluation
Takumi Ishiyama, Ryo Shibata, Gou Hosoya, Hiroyuki Yashima,
pp.492-492
Publication Date:2018/10/18
Online ISSN:2188-5079
DOI:10.34385/proc.55.We-AM-Poster.2
PDF download
Summary:
LT (Luby Transform) codes are the first realization of fountain codes [1]. LT codes employs belief-propagation (BP) algorithm to decode k input symbols from n received coded symbols. Ripple is defined as the set of coded symbols of degree 1 in the decoding process of LT codes. The decoding process would terminate if the Ripple diasppears. In this study, we propose a new degree distributions, which combines the Robust Soliton distribution (RSD) and the Heavy tail distribution [2] with the upper limit parameter D, to improve the performance of LT codes. By adjusting the degree upper limit parameter D in the Heavy tail distribution, we suppress the occurrence of high degree coded symbols and increase the Ripple in the decoding process. The simulation result shows that the proposed degree distributions has better decoding performance than the Robust Soliton distribution.