Distributed Modulation Classfication Based On Corrlated Cyclic Spectrum

The modulation recognition is not only a prerequisite for right signal demodulationbut also a key technology in un-cooperated communication, and it is widely applied forboth military and civil use. Due to the scarce priori knowledge about transmitted signalsand interference from environment, blind identification of modulation becomes reallydifficult.The unique statistical characteristic of digital signal as cyclostationary makessignals different from each other, and hence it has a lot of advantages to distinguish thesignals based on this characteristic. For examples, the clearance of cyclic spectrummakes it easy to tell the difference among signals; Absence of requirement in priorinformation would make it possible to do blind demodulation; it works well in low SNRcondition owing to noise interfering little to this unique statistical characteristic.The influence from sampling rate and baseband signal filtering, which would existin real communication, would finally lower correlated characteristic of cycle spectrum,intensify the difficulty of classification and recognition. In multipoint distributed system,cooperation identification based on information fusion would enhance the intension ofinformation, strengthen noise suppression. Main ways to realize Information fusion aredata fusion, characteristics fusion and decision fusion. Data fusion would collectinformation from points, and strengthen the amplitude of signal and reduction thestrength of noise. Characteristic fusion strengthen characteristic of signal. Decisionfusion makes decisions according to a particular axiom. These three fusions’ objection iselevating the recognition performance.This paper mainly does researches on six kinds of signal modulation recognition as2ASK、BPSK、QPSK、2FSK、4FSK、MSK to compare data fusion and characteristicfusion by analysis of signal power and cycle spectrum, classification and identificationof signal characteristics, the impact of selected characteristic on each parameterincluding in-band SNR, quantity of data, roll-off factor of baseband molding, simulationof signal effectiveness based on phase difference, delay inequality, difference in carrierfrequency and SNR. The results of simulations show out that data fusion would strengthen the amplitude of signal, and lower the strength of noise. Characteristic fusioncould increase the intension of signal characteristic. In the condition of same signalparameters multi-points’ cooperation recognition works better than one point, so as datafusion to characteristic fusion.