Automatic Modulation Recognition Based On Software Defined Radio

Modulation recognition of communication signals is one of the core technologiesof Software Radio (SR). On the premise that any modulation information is unknown,it can analyze the modulation types and estimate the modulation parameters of thecommunication signals from the received signals, which is helpful for subsequentsignal processing. How to efficiently monitor and recognize the modulation types isan important problem. Support Vector Machine has powerful ability in patternrecognition, and it is widely used in modulation recognition with its better stabilityand mistake tolerance.Firstly, this thesis mainly researches on the modulation recognition algorithmbased on SVM. The algorithm with instantaneous features is introduced, and featureextraction methods are analyzed. This paper has analyzed the characteristicparameters with the variation tendency of different signal-to-noise (SDR) bysimulation experiment. In order to classify AM、FM、2PSK、8PSK、2FSK、4FSK、16QAM and64QAM, this article does not directly use one-against-all (OAA),one-against-one (OAO) and binary tree (BT) in SVM multi-class classification, butpresents a method which is from Inter-class to Intra-class. The multi-classclassification is as much as possible divided into two classification problems and thenumber of classifiers is in decline. In simulation and comparison with the recognitioneffect of OAA, OAO and BT to classify digital modulation signals, the simulationresults show that BT is the best. In order to test the recognition performance of thismethod, Radical Basis Function (RBF) neural network was applied in modulationrecognition. The simulation results show that the recognition result of SVM is betterthan RBF.Secondly, A recognition method is proposed in this paper，which is based onmixed kernel function support vector machine and genetic algorithm (GA)，toimprove the classification ability of support vector machine. The method combinesthe learning performance of support vector machine with the searching performanceof genetic algorithm．Compared with traditional methods，this face recognitionmethod has the features of faster，less error and more efficient. Using the geneticalgorithm (GA) to optimize the parameter of SVM based on mixed kernel can givefull play to the advantages of various kinds of kernel function and the classification performance of SVM can be enhanced greatly.At last, the modulation recognition algorithm based on SVM is implemented onthe Digital wireless signal detector. In the actual environment test, AM、FM、2PSK、8PSK、2FSK、4FSK、16QAM and64QAM have a higher recognition rate.