Investigation Of LDPC Coded-Modulation Technique

This thesis studies the issue of efficient decoding algorithms of LDPC codes and the analysis and mapping design criterion issue of Low-Density Parity-Check (LDPC) coded-modulation. LDPC is a kind of liner block codes which has good performance .It was brought forward by Gallager in 1962 years. Because of computer's ability and related theory's weakness, the excellent code type was not thought much of in the scientific field. With the science and technology developed, its characteristic about high encoding gains, fast decoding, parallel decoding ,particularly, it has near Shannon limit performance on the condition of connecting with BP iteration decoding algorithm .All of this make people recognized the good performance of LDPC again and the research of LDPC come into a new phase ,and are very promising in many applications.With the development of the mobile communication to the high speed bandwidth, and the demand of transmission speed, frequency resource becomes more and more invaluable, in the future, communication system must make full use of limited frequency resource, the contradiction of frequency validity and power validity increasingly extrudes, the scheme of coded-modulation becomes the main method to overcome the contradiction. Coded-modulation is an error control technique which makes channel code and modulation become a whole, in order to obtain code gain, improve frequency validity and power validity of wireless channel transmission, we should consider associate LDPC with high order modulation.The paper first introduces the principles of Low density parity check codes, and encoding algorithm, elaborates decoding principle, introduces the sum-product decoding algorithm and log-domain decoding algorithm for binary LDPC codes. We compare the performance of LDPC with that of convolutional code, both theory and simulation can prove that LDPC puts up excellent performance with the BP decoding algorithm. We then give a discuss on the structures of LDPC coded-modulation systems, propose a model for coded modulation system of binary LDPC, use this model to joint LDPC with the coded-modulation ,and carry out the simulation for the performance in AWGN. With the accretion of modulation order ,the performance of decoding becomes bad ,but in the big SNR channel ,the LDPC of high order modulation data unfailingly transmits ,in the same time , ensures the data ratio improve several times. So unborn data communication, LDPC code which adopts high order modulation decoder could decrease decoding complexity and realize high speed data communication.On the base of researching BP decoding algorithms and coded- modulation system,we main analyze labeling map for LDPC-based coded modulation, and study the distributing of high order signal aggregate and labeling regulation for the bit to symbol, in order to design the optimum labeling map.Labeling map is the crucial design parameter for high frequency effective coded-modulation system. The performance of different labeling map could use Euclidean distance to describe, labeling map also decides output's soft information, affect decode performance. The fifth chapter detailedly discusses two optimum mapping criterions for high order coded-modulation, they are the harmonic mean of the minimum squared Euclidean distance and the constellation soft demodulation information. Make use of the harmonic mean of the minimum squared Euclidean distance as a design criterion to search the optimum labeling map, the bigger value of dh2 (u), the better performance of coded-modulation system. In the algorithm of constellation soft demodulation, we research the performance of soft decision applied in quadrature amplitude modulation (QAM) method ,especially16-QAM. Because belief propagation decoder needs the probability information of encoding sequence, uses soft demodulation algorithm to obtain logarithm likelihood ratio of every bit from high order modulation system. The bigger absolute value of LLR, the higher quality of the constellation soft demodulation information . With analysis of the simulation, if we want to make the absolute value of LLR big, we should consider border upon the same bit when design the constellation labeling.By combining both the two criterions, we propose a new mapping strategy. Results of research and simulation in this paper show that maximize harmonic mean of the minimum squared Euclidean distance and improve the quality of constellation soft demodulation information, the new designed signal labeling outperforms the Modified Set Partitioning Labeling and Random Labeling.With the demand of the high speed data transmission, propose a good code with high correct error ability——LDPC. Associating code with high order modulation will become one of the schemes for settling frequency resource, improving communication performance, so researching coded-modulation will have effective significance.