Research On Modulation Recognition Technique Of Multi-h CPM

Multi-h continuous phase modulation(Multi-h CPM)is a kind of nonlinear digital modulation signals with constant envelop and efficient use of bandwidth.It has broad application both in civilian and military communication field.It is quite significant to research the problem of its modulation classification.In this paper,two methods based on the extraction of signal feature are used to study the modulation classification of the signal of Multi-h CPM.Firstly,existing algorithms of modulation classification of digital communication signals are analyzed and their performances are compared.According to itself characteristics,cyclostationary feature-based and approximate entropy(ApEn)technology-based,two methods based on the extraction of signal feature,are chosen.Secondly,the method of cyclostationary-based is used to study the problem of intra-class recognition.The paper analyze the cyclostationary feature of some classical signals such as FSK、PSK e.g.and signals of CPM,and cyclic spectral of these signals are presented by the simulation,from where we can analyze the spectrum envelop and discrete spectrum lines.Furthermore,the cyclostationary feature of Multi-h CPM is analyzed according to the practical applications,and design a procedure of extracting the statistical feature.Then we make the intra-class recognition of signal set within Multi-h CPM by feature extraction of cyclic spectral.Finally,the method of approximate entropy-based is used to study the problem of within-class classification.Existing original classification procedure of ApEn focused on single modulation index CPM is learned to apply to the study of Multi-h CPM.In view of the limitation of the procedure,the paper proposed two improvements: one is that the preprocessing of the received signal can weaken the sensibility of the system to the noise,and improve the performance when the system is in low SNR region;the calculation of two-dimension feather of ApEn using “segmented-c” synthesize the advantage of different c value in different SNR region;the other is extracting the signal feature of instantaneous phase to build 3D-feature vector to apply more feature imformation into the classification system.The computer simulation shows that,the modified procedure makes better performance of classification system in the whole SNR region.