Blind Recognition of LDPC Codes Within a Candidate Set Based on Reliable Parity Checks

Blind recognition of channel codes has received a lot of attention in recent years. To improve recognition performance, some studies utilize reliable codewords instead of all received codewords for recognition. However, it is almost impossible to find reliable codewords when the code length is large and the signal-to-noise ratio is low. To address this problem, a novel blind recognition method based on Reliable Parity Checks (RPCs) is proposed in this paper. First, reliable bits in the received data are selected using a reliability threshold obtained from the simulation experiments. Then, only RPCs, i.e., the parity checks corresponding to these reliable bits, are used for recognition. Fortunately, this method is compatible with all existing blind recognition methods based on parity checks. Simulation results demonstrate that applying RPCs to the average log-likelihood ratio recognition method not only achieves a performance gain of about 1.5 dB, but also reduces computational complexity by over $90 \%$.