Modulation Recognition Technology Research For Digital Signal Based On Compressed Sensing

In digital wireless communication technology,modulation recognition at the receiving end is a key link in the process of signal demodulation and subsequent signal analysis.Only if the signal modulation mode is accurately identified,then subsequent signal processing tasks could be completed.Traditional Modulation recognition algorithm(Modulation Recognition,MR)are mostly based on Nyquist sampling theorem.However,with the development of communication technology,signal bandwidth and the amount of data also increase dramatically.So,the Nyquist sampling rate of the broadband signal will bring the ADC(Analog to Digital Converter,ADC)huge challenges,and the large amount of data for storage and transmission can also cause the rising of receiving cost.In this case,this thesis introduces the Compressed Sensing technique(Compressed Sensing,CS)to compress the received signal sampling,and uses the compressed data to reconstruct the high order statistic characteristics,so as to complete the identification of signal modulation mode.This thesis focuses on modulation recognition algorithm and compression perception theory.And a CS-HOC method is proposed,which is based on higher Order Cumulants(High-Order Cumulants,HOCs)and compression perception of modulation recognition algorithm.In CS-HOC algorithm,this thesis analyses signal sparsity for the undersampling signal on Walsh-Hadamard Transform(Walsh-Hadamard Transform,WHT).Therewith the undersampling linear relationship between vector and higher-order cumulant is established.So,at the actual receiving end,the higher-order cumulant features can be reconstructed based on undersampling and the signal modulation mode recognition can be completed.And in the deduced compression perception basic equation,the perceived matrix was a part Hadamard Matrix that owes the simple structure and fast operation advantage,which make the system complexity low.For mixed modulation signal,this thesis studies the relationship between characteristic extraction and signal synchronization and expands the CS-HOC method of single modulation signal into mixed modulation signal.Assume that the number of single modulation signal is known and implement timing synchronization and compressed samplings on mixed modulation signal.Based on the compressed samplings,the recognition feature is reconstructed,and the recognition feature is used as the sample parameters of SVM(Support Vector Machine,SVM)classifier.Similarly,at the actual receiving end,the predicting samples are get based on the compression-reconstruction process of received mixed modulation signal.At the same time,the MC task of mixed modulation signal can be completed through the trained SVM classifier.Simulations show that the CS-HOC algorithm could well classify single modulation signal and has a certain recognition correct rate for mixed modulation signal as well,which prove the validity and effectiveness of proposed scheme.