Research On Self-adaptive Satellite Communication System Modulation Decision And Modulation Recognition Technology

Satellite communication has become an important basic industry in thecommunication field. Adaptive satellite communication technology is proposed toweaken the satellites interference and to solve problems that different satellitescommunication systems were unable to exchange information.This system applies Maximum-Likelihood (ML) algorithm to estimate SNR.Simulation deduces the modulation mode switching threshold. In addition, theadaptive modulation switching threshold is the trade off between spectral efficiencyand Bit Error Rate (BER). Furthermore, the modulation mode decision strategyshould be considered to achieve the highest spectral efficiency and to maintainconstant average BER. The design of an integrated SNR estimation and adaptivemodulation mode decision system is highly challenging. Numerical results showthat the design fulfills superior performance.Then based on the adaptive satellite communication modulation decisiontechnology, this paper proposes a new feature selection based on resistance to thenoise analysis of feature parameters. Focuses on several classic feature parametersin satellite communication dynamic SNR environments for the robustness ofgaussian white noise. Through simulation experiment, under ten kinds ofmodulation signal, different characteristic parameters change with SNR, thevariance values are presented. Variance values can reflect the noise robustness ofcharacteristic parameters. After the selection two characteristic parameter sets areachieved. Further experiments, estimated SNR range sequence is divided into foursections, received signal samples are separated into four sections and four SVMsare set for modulation mode recognition experiment. And the two test signal groupsare labeled respectively by the two feature sets before or after selection.Experimental results show that four periods of SNR range, the sample identificationcorrect results with selected feature is higher than unselected feature. It is verifiedthat the feature set after selection has more stability with noise changes. Simulationresults show that the feature selection method in this paper not only improves theclassifier to strong sample generalization ability. Also improves the work efficiencyand accuracy, has great practical significance.