The Research On Automatic Modulation Classification Based On Grey Wolf Optimizer Least Square Support Vector Machine

Automatic modulation classification(AMC)means a series of processing on modulated signals,which includes extracting features and classifying signals,under the condition that received signal is unknown.With the galloping development of digital modulation,AMC,which has been palying an important role in the area of wireless communication,has become a hot research spot because of its military significant and civilian value.AMC is one of the appilication scenario of machine learning and intelligent information processing.Researches on AMC can be definitely expanded to other problem of intelligent information processing,which has broad prospects.Based on above background and analyse problem existing in researches on AMC,this paper focuses on the recognition of MPSK and MQAM under AWGN channel.A new feature extraction method of modulated signal based on high order cumulants and local mean decomposition is proposed.Besides,this paper designs a new modulation classifier based on cuckoo search algorithm and grey wolf optimizer.The major work in this paper is summarized as below:1.Against the problem that high order cumulant cannot effectively differ MQAM from MPSK and EMD has serious endpoint effect,approximate entropy of LMD is combined and a new feature extraction method based on HOC and LMD is proposed.The method extracts feature by calculating 2/4/6 order cumulants of signals and the first two approximate entropies of the sequences after decomposition.2.Traditional least square support vector machine has no ability of changing its hyper parameters,and most swarm intelligence algprithms are known with high computational cost and low robust.A modulation recognition method based on least square support vector machine whose hyper parameters are optimized by grey wolf optimizer algorithm is proposed.The simulation results show that the recognition rate is 94.1% when SNR is-3 dB,as well as over 99% when SNR is 6 dB.Besides,the method proposed is approved to bring about self-adaption of hyper parameter of LSSVM and reduce the iterations.3.Grey wolf optimizer is an algorithm that is apt to fall into local optimal solution.Against this defect and based on aggressive reproduction strategy and Levy flight of cuckoo search algorithm,this paper improve grey wolf optimizer with cuckoo search algoritm.After location update of grey wolves,improved algorithm will search location for second time to expand search space.The simulation results show that the recognition rate is 96.7% when SNR is-3 dB,as well as over 99% when SNR is 0 dB,which is a better result than that of GWO-LSSVM.