Efficient-memory and low-latency BP decoding algorithm for polar codes

In order to reduce memory and latency of the belief propagation (BP) decoding algorithm, we use new constituent codes to simplify the message passing of factor graph for polar codes. Since the left-to-right (right) messages of top level of constituent code are directly updated by the right-to-left (left) messages without visiting its sub-factor graph, which saves the corresponding memory entries and decoding latency. Therefore, based on our proposed constituent codes, we can obtain a memory-efficient and low-latency BP decoding algorithm. Numerical results show that the proposed decoding algorithm saves about 4.6%-17.6% memory entries of the XJ-BP decoding and the latency is reduced by 19%-23.5%.