Researches On Typical Signals’ Modulation Type Recognition Methods

Modulation type recognition is one of the important application fields of modernno-cooperative communication systems. The aim of performing modulation recognitionis to identify the modulation type of communication signals by some means. Suchtechnique can provide information for non-partners, who would take some measuressuch as interception, interference and threat analysis.Existing algorithms don’t cover enough scales, and most of them are not practicalenough. Against the two problems, the classification of typical communication signalswas deeply researched in this article. The algorithms to identify the modulation types of21kinds of analogue and digital communication signals were analyzed and verified, andthe joint classification methods were proposed. The main contents and work of thisdissertation are as follows:The algorithms for SNR estimation, baud rate estimation and timing error estimationwere studied, which are suitable for the objects of this thesis, and the simulation resultsand analysis were given. An improved baud rate estimation algorithm which could beused for ASK and QAM signals was proposed. Simulation results showed theeffectiveness of the algorithms mentioned above.The intra-class recognition algorithms for digital signals were deeply studied. Forcommenly used FSK, the number of squared signals’ spectrum peaks was utilized; for6kinds of PSK signals, a modified identification scheme based on different spectrumswas put forward; for ASK and QAM, a practical algorithm was proposed, that is,performing DDC and match filter to received signal, using the estimation result of theparameter and interpolation algorithm to get symbol synchronization automatically,employing synchronization extraction and then drawing the numbers of symbolenvelopes to confirm modulation orders by clustering algorithms.Fully using the existing parameters, and performing modification to some of them,we presented a joint recognition scheme for these21kinds of signals. This scheme onlyrequires the rough estimation of carrier frequency and band width, need no other prioriinformation. In the process of designing, the influences of shaping filter, frequency offset and phase offset was taken into consideration, and the generation of signal wascloser to actual situation. Especially, we employed SNR estimation algorithm to setadjustable threshold when classifying PSK and QAM signals. This work improved theperformance of inter-class classification in relatively low SNR condition. As simulationresults show, under the condition of SNR is no less than7dB, the correct classificationrate is over90%for most signals; on the other hand, when SNR is no less than11dB,the correct classification rate is over90%for all signals. Moreover, the performance ofthis procedure is insensitive to the changing of modulation parameters, showing thefeasibility and practicality of the proposed method. And last, by using the simulationtool, the modulation software was designed and demonstrated.The study in this thesis extended the scope of the existing identification methods,enriched the exsiting recognition technology, and has been employed in a project,reaching the requirements of which perfectly.