Modulation Recognition For Communication Signals Based On Sparse Representation Classification

Modulation recognition is a technology that recognizes the type of the receivedsignal at the receiver. Now the popular classifiers are decision-tree classifier and neuralnetwork classifier, etc. Decision-tree classifier performance depends on the decisionthreshold set, lead to great changes have taken place with the decision thresholddetermination of recognition rate. Neural network classifier has the problems of owelearning, learning and fall into local minimum, etc. Aiming at the problems of theexisting classifiers, in this paper, the sparse representation algorithm is applied to themodulation recognition, by using the algorithm based on the sparse representationclassification recognition, the classifier can not only avoid existing classifier problems,but also has strong robustness and anti-noise performance is strong, the characteristicsof the recognition rate is higher.In this paper, my work and achievements are as follows:First, research the theory of sparse representation, discuss the property of the sparsenature, and introduce the algorithm of dictionary learning and the theory of sparserepresentation classification. Based on the above content, this paper puts forward animproved classifier of the sparse representation based on dictionary learning algorithm,which based on ClassDKSVD according to the dictionary class learning method. Andthen this algorithm is combined with sparse representation classifier thought as theclassifier recognition system, the communication signals modulation mode recognition.Finally on the basis of the algorithm, the recognition algorithm based on instantaneousfeature and recognition algorithm based on higher-order cumulant is studied.Signal information is included in the amplitude, frequency and phase of themodulated signal, therefore in recognition algorithm based on instantaneouscharacteristics, this paper chose five instantaneous characteristics combined withClassDKSVD algorithm proposed in this paper, seven kinds of signal,2ASK、4ASK、2FSK、4FSK、2PSK、4PSK、16QAM are recognized. The experimental simulationresults show that the proposed algorithm is to identify many type, the correctrecognition rate is high.In the recognition algorithm based on high-order cumulant study, this paper chosefour high-order cumulant characteristics combined with ClassDKSVD algorithmproposed in this paper, four kinds of signal,4QAM、8QAM、16QAM、64QAM are recognized. The experimental simulation results show that the proposed algorithm is toidentify many type, the correct recognition rate is high. It has the validity andsuperiority.