Research On Blind Recognition Technology Of LDPC Codes

In the scenario of non-cooperative communication,obtaining the technical parameters,work contents and other intelligence information from the intercepted data has important military significance and intelligence value,during this process,the technology of blind recognition of channel coding plays a key role.The main work of this technology is to estimate both the type and the parameters of channel coding about communication side under the condition of knowing no or only partial prior knowledge by using mathematical tools and algorithms accordinig to the intercepted data,which prepares for the next decoding work.In this thesis,we studied the technology of blind recognition of Low Density Parity Check(LDPC)codes.In general,the length of practical LDPC codes are long and their parity check matrices are sparse,which make traditional recognition algorithm hard to implement in the noisy environment,so,related research results are rather rare.At present,the researches on the recognition of LDPC codes are mainly carried out under the concept of a candidate set,whose main work is to find the most likely source of intercepted data in a constructed LDPC code parameter library.The blind recognition of LDPC codes under a candidate set weakens the problem of parameter estimation in the the galois field of two elements into a hypothesis testing problem of the galois field of two elements,it needs a lot of data and priori knowledge as support,which are difficult to meet in the background of non-cooperative communication scenario.Aiming at this problem,we put forward the algorithm of blind recognition of LDPC codes without a candidate set,which trys to blindly estimate code length and synchronization,code rate and linear constraint relation and eventually realize the blind recognition of the sparse parity check matrix according to the intercepted bitstream.For the recognition of code length and synchronization,most of the existing algorithms are based on rank criterion,whose recognition right rate are not ideal under high bit error rate and these algorithms have a high complexity.As for the code rate and the parity check matrix,the method of linear matrix analysis is sensitive to error,the hadmard transform method has a high complexity,in addition,the above methods can not make the parity check matrix sparse.Aiming at the deficiencies of these traditional recognition algorithms,combining the sparse characteristics of LDPC codes' parity check matrix and starting from the dual space perspective of codewords,an improved method based on solving low-weight linear combination of columns is proposed to obtain the dual codes and based on this method,an algorithm based on dual space is proposed in this thesis to solve the blind recognition of code length,which makes the right rate of the code length higher than traditional rank criterion method,then we use the method of sloving the support interval of the dual vectors to speed up the searching of synchronization,next,according to the code length and synchronization we identified,an iterative screening algorithm is designed in combination with the above improved method of solving low-weight linear combination of columns to identify the linear constraint relationships and the code rate.Finally,we propose a sparse algorithm to reconstruct sparse parity check matrix of LDPC codes and eventually we realize the blind recognition of LDPC codes without a candidate set.