Detection And Identification Of Wireless Multimode Signals Based On SVM Mine Environment

Mine informationization is an effective guarantee for coal mine safety production.Many subsystems of mine information often adopt different signal modes.To form an integrated information system,multi-system integration must be achieved.Detection and recognition of multi-mode signals is the key to multi-system integration.In this thesis,the influence of mine environment on signal characteristic parameters is studied.Support vector machine(SVM)is used as classifier to establish classification and recognition model of mine environment multi-mode signals,which provides theoretical basis for the research and fusion of underground multi-system.In view of the mine environment,the modulation recognition of signals is studied,and the pattern recognition method is selected,which is divided into two parts:feature parameter extraction and classifier design.Aiming at the problem of feature parameter extraction,the fourth-order cumulants of.signals are selected as feature parameters,and the second-order moments,fourth-order moments and fourth-order cumulants of signals in white Gaussian noise channel are analyzed and obtained.On this basis,the relationship between the fourth-order cumulants of signals and shadow fading and small-sc.ale fading is further analyzed,and the expressions of the four-order cumulants through these two fading channels are obtained.Aiming at the problem of classifier design,SVM is selected as the target classifier,and the fourth-order cumulant of signal is used as the input of SVM.In the case of binary tree classifier,one-against-rest classifier,one-to-one classifier and decision tree classifier,the classification and recognition of BPSK,OFDM,16QAM and 64QAM signals are realized respectively.The simulation results show that the recognition effect of decision tree classifier is more than one.The performance of the three classifiers lbased on SVM algorithm is comparable,and the recognition effect is not ideal at low SNR.signal-to-noise ratio(SNR),a method of optimizing SVM classification and recognition is proposed.The data sample set is divided into test data set and training data set.Particle swarm optimization and genetic algorithm are used to optimize the penalty factor and kernel function of SVM in training data set.The optimized SVM model is obtained and used to test and classify the test set.The simulation results show that the average recognition rate of the four signals can reach more than 80%in the three channel environments with signal-to-noise ratio of-5 dB,and that the average recognition rate of the four signals can reach more than 90%in the three channel environments with signal-to-noise ratio of-3 dB.