Research On Blind Identification Algorithms Of Linear Block Code Parameters

Linear block code is a kind of channel codes with excellent performance and have been widely used.At present,most blind identification methods for linear block code parameters are only for a special type of linear block code(such as BCH code,RS code,LDPC code,etc.).In order to improve the application range and fault tolerance of the blind identification method of linear block code parameters,we identify the linear block code parameters under the premise that the received data are encoded with a kind of linear block code,but the specific encoding type and other information are unknown.The code length recognition and the parity check matrix reconstruction are studied in this paper.In terms of code length recognition,two traditional code length recognition algorithms are studied and simulated,which are rank criterion method and code weight analysis method.Furthermore,the application range,advantages and disadvantages of the two algorithms are summarized.Based on the two method,this paper establishes the concept of "normalized column weight vector" and proposes a code length recognition method based on the cosine similarity of normalized column weight vector.The concept of “normalized column weight vector” is defined as the ratio of "1" in each column of the matrix.The code length identification is achieved by measuring the difference between the normalized column weight vector of code word matrix after Gaussian column cancellation and the one of random binary matrix by using cosine similarity.The simulation results show that in the code length recognition of C(15,7)and C(31,6),the fault tolerance of the proposed method is improved by 83.33% and 50%as compared with that of the code weight analysis method,respectively.Furthermore,the proposed algorithm has good fault tolerance for linear block codes with different code lengths and code rates,which can be widely applied.In terms of the parity check matrix reconstruction,the theory of Gaussian elimination method and Walsh Hadamard transformation method are studied,and the Walsh Hadamard transformation method is simulated.Different from the traditional parity check matrix reconstruction algorithms which are based on matrix theory,this paper reconstructs the parity check matrix by searching the linear constraint relation between the parity check bits and the information bits of the linear block code.By introducing the concept of association rule mining in the field of data mining,this paper proposes the reconstruction algorithm of parity check matrix based on association rule mining.This paper carries out association rule mining on the codeword matrix to find the possible constraint relation,and finds the correct constraint relation by the inherent properties of linear block codes,then the parity check matrix could be reconstructed.The simulation results show that in the parity check matrix reconstruction of C(7,4)and C(15 7,),the fault tolerance of association rule mining method is improved by368.75% and 221.43% as compared with Walsh Hadamard transform method,respectively.Therefore,the proposed method has excellent fault tolerance.