Detection And Identification Technologies For Communication Signals Based On Goodness-of-Fit Test

Aiming at solving the communication reconnaissance signal processing problems with only a few samples,signal detection,modulation classification and SNR estimation based on Goodness-of-Fit(GoF)test are researched in this dissertation.Owing to the fact that it bases its theory directly on the distribution function,the GoF test based algorithms can give satisfied performance with few samples.The main research work and achievements of the dissertation are as follows:1.Summarized the existing GoF tests in statistics,and presented the advantagesand shortcomings for each,which can be served as basis for the selection ofsuitable test used in communication signal processing.2.The common procedure for signal detection based on GoF test is summaried.The algorithms are analysed with different decision statistics,test statistics andone-sample or two-sample scenario.Its performances are simulated underdifferent signal models.3.Signal detection algorithm with noise power uncertainty based on compositeGoF test is proposed,which has nearly the same performance as that for simpleGoF test based detector but without a prior information about the noise power.The new detector nearly introduces no extra computational load.4.The common procedure for modulation classification based on GoF test issummaried.The algorithms are analysed with different decision statistics,teststatistics and one-sample or two-sample scenario.Its performances under flatfading environment are simulated.5.The procedure for SNR estimation based on GoF test is presented,and twocomplexity reducing schemes are proposed by exploring the ordercharacteristic of the SNR matching pool.Comparison with the Method-of-Moment(MoM)based estimator indicates its superiority with few samples.Thecomplexity reducing schemes' effectiveness are also verified.6.A joint estimator of modulation format and SNR based on KS test is proposed.The joint estimator can obtain the estimates of modulation format and SNRvalue in a unified procedure using limited signal samples.Analysis andsimulation results have verified its effectiveness.