| Channel coding is an important step in the communication system,and channel coding data identification technology contribute important research value to the communication field.The channel coding data studied in this thesis are respectively "Pseudo-random scrambling coding data","Block coding data" and "Convolutional coding data".Different from the traditional channel coding data identification technology,this thesis mainly studies the channel coding identification technology based on convolutional neural network.The main work includes three aspects: the original data transformation processing method,the construction of multiple error coding data sets and the improvement test of deep learning model selection.In this thesis,the error correction coding recognition technology based on deep learning method solves the problem that only single coding can be identified in the traditional technology recognition.Secondly,the method proposed in this thesis can identify more accurately under high error code.Finally,based on the generated channel coding data set,VGGNet,RerNet and IncptionNet are used in the case of10 ‰,5 ‰,and 1 ‰ error respectively,Three kinds of convolutional neural networks are used to identify the data set,and the recognition rate can reach more than 95%under the condition of 1 ‰ error.In this thesis,the recognition effect,recognition efficiency,model visualization and other aspects of the recognition algorithm are tested,and different scenarios applicable to different algorithms are summarized.In conclusion,the research in this thesis fully demonstrates the feasibility and advantages of channel coding data recognition based on deep learning,and discusses the applicability of deep learning in the field of channel coding.Compared with the traditional research,the channel coding and identification technology studied in this thesis can identify higher error data,and it is more intelligent,automatic and efficient in operation. |
PDF Full Text Request