Research Of Blind Recognition Methods Of Error Correction Code

Blind recognition of error correction codes is an important research topic in the field of communication countermeasures,which aims to fully mine the differences between different error correction codes without any prior information,identify the relevant information of error correction codes,and provide technical support for further recognition of communication signal features.The problem of blind recognition of error-correcting codes mainly includes code type recognition and code parameter blind recognition.The existing research mainly focuses on the recognition analysis of code parameters based on the known code type of error-correcting codes.This thesis mainly focuses on the problem of code type recognition of error correction codes under completely blind conditions.Aiming at the problem of poor fault tolerance of current algorithms based on matrix rank property,a blind code type recognition algorithm based on depth distribution property is proposed to realize the recognition and classification of linear block codes,convolution codes and Turbo codes.In order to further improve the universality and anti-noise performance of the code type blind recognition algorithm,an error-correcting code type blind recognition algorithm based on Recurrent Neural Network(RNN)is proposed,which realizes the blind recognition of various code types through Neural Network’s powerful learning ability.Finally,combining the characteristics of the code word and the theory of multiple domains,a code type recognition method based on the improved Euclidean algorithm is proposed.Bose-chaudhuri-Hocquenghem(BCH)and Reed-Solomon(RS)codes,which are widely used in satellite communication,are analyzed for code type and parameter recognition.This thesis focuses on the study of error-correcting code type recognition,and the main research work is as follows:1.Aiming at the problem that the traditional code type recognition analysis algorithm based on the rank characteristics of code word matrix is not fault-tolerant,a blind recognition algorithm based on the depth distribution characteristics of code type recognition is proposed to complete the code type blind recognition analysis of linear block codes,convolution codes and Turbo codes.Firstly,the improved Gaussian elimination method is used to solve the starting point and code length position of linear block codes,and the corresponding generation matrix is solved by using the depth distribution of linear block codes,so as to complete the blind recognition analysis of linear block codes.Then,different code types are uniformly equivalent to linear block codes by means of code type reconstruction,and blind recognition analysis of convolutional codes and Turbo codes based on depth distribution characteristics is completed,which further improves the scope of application of the algorithm.Simulation results verify the effectiveness of the proposed algorithm.Compared with the problem of insufficient anti-noise performance of traditional matrix rank estimation algorithm,the algorithm effectively improves the fault tolerance performance of recognition.2,aiming at the error correction coding type blind recognition algorithm mainly concentrated in the field of mathematical analysis of a specific type recognition,the universal existence deficiency and the problem of insufficient anti-noise performance,to carry out the error correction coding based on the theory of the deep learning type blind recognition research,put forward the base cycle error correction coding type blind recognition algorithm of neural network,and carry on the improvement of parameters initialization algorithm optimization.Compared with traditional Gaussian initialization and Xavier initialization algorithms,the weight initialization algorithm adopted in this thesis initializes the self-connected weight of RNN hidden layer,effectively controls the influence of the state of the last time step on the current time step,inhibits the disappearance of gradient,reduces the complexity of the algorithm,and effectively improves the training efficiency.The simulation results show that the blind recognition algorithm based on improved RNN improves the anti-noise performance significantly,has high recognition rate,and has good practicability.3.On the basis of the classification of error correction code type,the BCH code and RS code,which belongs to the same category of cyclic code,are further classified and recognized,and the blind recognition of error correction code type based on the improved Euclidean algorithm is proposed.Euclidean algorithm is mainly aimed at solving the largest common factor between code words.The algorithm is mainly aimed at identifying and analyzing the circular code in the linear block code.By obtaining the original polynomial of the code word,further obtaining the generating polynomial,so as to complete the recognition.To solve the problem of parameter recognition of BCH codes,firstly,the maximum common factor of the BCH code word matrix is found by the toss and turn division method,and the decoding length is calculated by calculating the distribution of the maximum common factor coefficient.Moreover,a new evaluation index based on information entropy is proposed,which is insufficient to distinguish the discrete differences between data compared with the traditional evaluation index.The index proposed in this thesis improves the fault-tolerant performance of the algorithm effectively.Aiming at RS code recognition problem,on the basis of finding the maximum common factor through the division method,the mean square error index is proposed as the decision criterion,which can effectively improve the fault-tolerant performance of the algorithm and realize RS code recognition and analysis under the condition of high ber.The effectiveness and engineering practicability of the algorithm are verified by the actual data recognition analysis.