LDPC碼之混和解碼法-MP法與A*演算法; On the Hybrid Decoding Method for LDPC Code by Using the Message Passing and the A* Algorithms

楊旭弘, Hsu-Hung Yang

LDPC code used by the advanced communication standard of the next generation is an error control code. Its error correction ability may approach the Shannon’s theoretical value. With the MP algorithm, it can decode received samples in high speed from transmitter. However, the MP method is suboptimum and optimum solution is not guaranteed. In this thesis, a hybrid decoding method is formed by combing the MP and the A* methods. We first make comparison between the MP and A* method and then show their combined performance. From the results, the coding gain of A* model is higher than that of MP model. The decoding complexity increases exponentially with the codeword length for A* method, but the MP method assumes linear increase. Using the hybrid structure, its coding gain is the same as that of A* method. Moreover, decoding complexity is slightly greater than that of MP one. For codeword length 96 bits and coding rate 1/2 and BER = 10E-5, the hybrid structure outperforms 1.4 dB coding gain than the MP method. This improvement only required 1% received sequences to be sent to A* decoding block. Our newly designed hybrid structure can solve both of the high coding gain and low decoding complexity while it has the ability to yield the optimum solution.

許恭睿, Kung-jui Hsu

Chia‐Lu Ho, Hsu-Hung Yang

Because the message passing (MP) method is suboptimum and optimum solutions are not guaranteed, we present a hybrid decoding method by combing the MP and the A* methods. We first make comparison between the MP and A* method and then show their combined performance. Using the hybrid structure, its decoding gain is the same as that of A* method. Moreover, decoding complexity is slightly greater than that of MP one. For codeword length 96 bits and coding rate 1/2 and BER = 10-5, the hybrid structure outperforms 1.4 dB decoding gain than the MP method. This improvement only required 1% received sequences to be transferred to A* decoding block. Our newly designed hybrid structure can solve both of the high decoding gain and low decoding complexity while it has the ability to yield the optimum solution.

Ligter, Luca

Saleh Usman, Mohammad M. Mansour, Ali Chehab

This paper investigates efficient decoding algorithms for LDPC codes. Alternating column-row message-passing (ACRMP) and Interlaced column-row message-passing (I-CRMP) schedules for decoding of LDPC codes are proposed and investigated in this work. Existing serial scheduling schemes for LDPC decoding are based either on column message-passing (MP) or row MP, and roughly converge twice as fast as Gallager's flooding-based MP schedule at high signal-to-noise ratio (SNR). To further accelerate the convergence speed of serial decoders, hybrid column-row MP schedules that perform multiple message passes between check and variable nodes within or between iterations are proposed. Our proposed I-CRMP schedule converges in less than half the number of iterations compared to the best existing serial decoding schedules. Compared to column MP, the added complexity of this scheme is proportional only to check-node-degree times more additions at variable nodes. This increase in complexity is moderate compared to the convergence acceleration factor that the scheme achieves. Superior performance of the proposed I-CRMP scheme is confirmed by decoding randomly generated as well as IEEE 802.11n/ac LDPC codes.

Tobias Lindstrøm Jensen, Nano Thaore, Shashi Kant