Research On Modulation Classification And Parameter Estimation Of Communication Signals

Modulation classification and parameter estimation for communication signals arereferred to judging the modulation type and estimating parameters of an interceptedunknown communication signal，which belong to pattern recognition and parameterestimation. They are of significant application values and broad practical prospect inboth civil and military communication fields, such as communication reconnaissance,spectrum monitoring and so on.In this dersertation, Intrinsic Time-scale Decomposition(ITD) algorithm andmulti-fractal theory are employed as the two main analysis tools, while the modulationclassification and parameter estimation are processed based on signal waveform inintermediate frequency, it has well taken advantages of the high time frequencyresolution and low computational complexity of ITD algorithm as well as thecharacteristic of insensitivity to noise of multi-fractal features. The author’s maincontributions are as follows:1. A modulation recognizer scheme based on time domain statistical features andSupport Vector Machine(SVM) is proposed. In the scheme, the ITD algorithm isemployed to extract instantaneous amplitude, instantaneous phase and instantaneousfrequency. A12-dimensional original feature vector is formed based on statisticalcharacteristics of these instantaneous information. Then, Fisher analysis is used toreduce the dimension of the original vector. Finally, the lower dimension feature vectoris input to a multiclass SVM to classify single carrier digital modulation signals,including7types of digital modulation such as2ASK,4ASK,2PSK,4PSK,16QAM,2FSK and4FSK. The ITD-based features can be directly obtained by means ofwaveform analysis, which not only has very low computational complexity, but alsohave good discriminability because of dimension reduction based on Fisher analysis.Simulations show that the modulation recognizer can provide high recognition accuracywith very low processing complexity.2. A modulation classification algorithm for multicarrier OFDM signal and singlecarrier modulation signals is proposed based on two multi-fractal features combinedwith a decision tree classifier. Firstly, singularity exponent corresponding to themaximum multi-fractal value from intermediate frequency waveform signals is regardedas the first feature, which is used to discriminate the signal set {OFDM,4PSK,16QAM}from the signal set {2ASK,4ASK,2FSK,4FSK}. Then, the spectrum span between multi-fractal points whose corresponding slope are respectively±1from the amplitudeof twice Fourier transform of squared intermediate frequency waveform analytic signalsis regarded as the second feature, which is employed to discriminate the signal set{OFDM} from the signal set {4PSK,16QAM}. Finally, a decision tree classifier isintroduced to classify these signals and the classification thresholds of each feature aredefined by empirical simulation. Numerical results show that this algorithm is robust tothe number of subcarriers and the length of the cyclic prefix of OFDM signal,raised-cosine shaping roll-off factor of single carrier signals and the signal to noiseratio(SNR), and that it can recognize OFDM signal in a wide SNR range with a highaverage correct classification probability.3. An algorithm for estimating carrier frequency and chip rate and a waveformbased algorithm for estimating pseudo-noise(PN) code of a direct sequence spreadspectrum(DSSS) signal are proposed, both of which are based on ITD algorithm. In theformer algorithm, instantaneous amplitude of the DSSS signal at a given frequencypoint is regarded as the analysis parameter. Chip rate and carrier frequency areestimated according to coarse search and elaborate search in DSSS signal frequencydomain separately. Firstly, instantaneous amplitude of analysis signals and incidentalsignals from several PN periods are cumulated. Then, differential calculation of theabove two cumulated amplitudes is used to remove the noise effect. Finally, chip rateand carrier frequency is estimated by line spectrum analysis to the spectrum of thedifferential signal. This algorithm appears to be simpler in implementation, faster incalculation compared with the traditional algorithms. In the latter algorithm, takingadvantage of high time-frequency resolution and convenience of real-time processing inITD algorithm, it directly analyzes fluctuation characteristics of differential signalbetween cumulative amplitudes from carrier frequency point and the first upzero-crosssing frequency point to find the positions of polarity alternating pointsbetween two adjacent chips in a period of PN code and to reveal the PN code，withoutnecessity of guessing algebra structure, where coherent accumulation concept is appliedto raise SNR. The algorithm has some advantages compared with the existingalgorithms for PN code estimation. Computer simulation validates the feasibility of theproposed algorithm.4. An ITD based algorithm for hop rate estimation of frequency hopping signal isproposed. Frequency hopping signal is iteratively decomposed by ITD algorithm and anew analysis sequence is derived which is composed by the maxima from theinstantaneous amplitude envelop of each layer proper rotation component. Hopping rate can be estimated by line spectrum analysis to the spectrum of this analysis sequence.Computational complexity analysis and numerical results show that the algorithm has alower computional complexity and proved to be faster and more effective comparedwith the empirical mode decomposition (EMD) based algorithm.