Reduction of Decoding Iterations for Zigzag Decodable Fountain Codes
T. Nozaki · International Symposium on Information Theory and its Applications · 2016
Fountain codes are erasure correcting codes realizing reliable communication on multicasting. In a previous work, we proposed a fountain code, named zigzag decodable fountain (ZDF) code, and showed that the ZDF code outperforms Raptor codes in term of the decoding performance. However, the conventional decoding algorithm for the ZDF code, which is regarded as a combination of packet-wise peeling algorithm and bit-wise peeling algorithm, requires a large number of iterations. The main purpose of this research is the reduction of the number of decoding iterations for the ZDF codes. The tree-structure expectation propagation (TEP) is a decoding algorithm for the low-density parity-check codes over the binary erasure channels and an extension of peeling algorithm. In this paper, we extend the TEP for the ZDF codes and propose a decoding algorithm combined the packet-wise TEP with bit-wise peeling algorithm. Simulation results show that the proposed decoding algorithm reduces the number of decoding iterations as compared with the previous decoding algorithm without loss of decoding performance.