A rateless encoder with adaptive symbol removal for AWGN and fading channels

Abstract

In this paper, we propose a novel symbols removal encoding (SRE) raptor code over binary additional white gaussian noise (BIAWGN) channel and fading channel. By calculating the theoretical logarithmic likelihood ratios (LLR) messages that each input symbol accumulates in the encoding process and removing those input symbols whose LLR messages are enough for successful decoding from the latter encoding process in the SRE scheme, the LLR messages from the channel can be allocated more properly among the input symbols and thus the SRE Raptor code can achieve better decoding performance than conventional Raptor code. By analyzing raptor code based on extrinsic information transfer charts (EXIT) under Gaussian approximation, we propose the method to calculate the theoretical LLR messages of each input symbol and design the proper threshold to remove the symbols. Finally the simulation results show that the SRE Raptor code can achieve better decoding performance obviously both over BIAWGN channel and fading channel compared with conventional raptor code and the equal degree encoding (EDE) Raptor code.