A Class of Low-Complexity Codes Based on Doubly Recursive Block Markov Superposition Transmission

In this paper, we introduce the doubly recursive block Markov superposition transmission (DrBMST) of short code. An important characteristic of DrBMST codes is that the degrees of the constraint nodes in their normal graphical realizations are at most three. As a result, DrBMST codes can be decoded with low complexity. We propose to use an enlarged code ensemble to analyze the performance of DrBMST under windowed maximum likelihood decoding. Further, the extrinsic information transfer (EXIT) chart analysis is used to study the iterative decoding thresholds of DrBMST code ensembles. The EXIT analysis shows that the iterative decoding thresholds of DrBMST code ensembles are comparable to those of the BMST codes. We have also compared the error performance and the decoding complexity of finite-length DrBMST codes with regular spatial-coupled low-density parity-check (SC-LDPC) codes under equal decoding latency. The comparison results show that the DrBMST code performs about 0.1 dB better than a (4, 8)-regular SC-LDPC code, but with lower computational complexity.