Signal Parameter Estimation And Modulation Recognition Based On Cyclic Spectrum

Modulation recognition and modulation parameters estimation of communication signals are of important significance for military electronic countermeasure and civilian spectrum management. At low SNR circumstance, traditional methods of parameters estimation and modulation recognition make it difficult to adapt to wireless communication environment which owns complexity and unpredictability.Meanwhile,signal processing based on cyclic spectrum analysis has a good performance on restraining noise and has been widely applied to the parameters estimation, feature parameter extraction and other fields. Considering the huge problems to be solved on cyclic spectrum analysis, we absorbed latest ideas and research products and to realize the modulation identification and parameters estimation with some new algorithms in this paper.Firstly, we analyzed some general cyclic spectrum period grams of digital and analog modulating signals. Based on the frequency domain smoothing algorithm, the paper analysis the cyclic spectrum estimation, which is influenced by the signal carrier frequency, symbol rate and the parity of the frequency resolution. And we proposed the joint section search algorithm. Through frequency multiplication and square transformation, the high order MPSK signal will cause the problem of the random noise. In order to solve it, this paper puts forward signal preprocessing process based on Hilbert transform . Furthermore,through mathematical deduction and simulation, we analyzed the classification of the parameters of the modulation signal on the spectral feature section and the modulation signal recognition process based on the cyclic spectrum analysis. The paper also discussed the matrix extraction algorithm and introduced the resemblance coefficient for a feature parameter suitable for low SNR environment. The simulation results show the algorithm owns a better recognition performance in low SNR condition. Finally, two main factors influencing the matrix extraction algorithm extraction coefficient and the point of discrete Fourier transform are analyzed. The simulation proved that a relative low extraction coefficient and a relative more point of discrete Fourier transform show a better recognition performance.