Study On Modulation Recognition And Parameter Estimation Based On Sparse Reconstruction

Radar reconnaissance is an important area of radar countermeasures. Intercepting enemy radar signal and signal detection are the primary mission of radar surveillance. Traditional interception technology and detecting signal are based on Nyquist Sampling Theorem, while compressive sensing(CS) theory breaks this limitation. CS can use the sparsity of signal and intercept small amounts of observational data to achieve those goals.Thesis applies the sparse reconstruction thought to modulation pattern recognition and parameter estimation. We can still extract information from the original signal by small amounts of data. Because it is the theoretical basis of research, we select only the most basic three radar signals for analysis. Specific content can be summarized in the following two sections:The first part studies the intra-pulse modulation pattern recognition based on sparse reconstruction problem. First, we demonstrate that the three of unknown radar signal has a premise condition that they can be sparse reconstruction. Through observation matrix and sparse matrix consisting of a dictionary, we map a small number of observation signals to frequency domain and use reconstruction algorithm to determine the sparse coefficient. Comparing with sparse coefficient and corresponding relationship of pulse width, to achieve the goal of signal modulation pattern recognition. Then through the theoretical analysis and simulation experiments, we discuss the correctness of the method and explore the recognition success rate under different compression ratio and how to select the threshold under a certain compression rate and discrimination rule. Adding the signal to noise, we simulation analysis the influence of noise on recognition rate. Finally, it is concluded that, when selecting the appropriate threshold, this method can be applied to the modulation pattern recognition, and when the signal-to-noise ratio is not particularly low, has the sound recognition success rate.The second part studies the typical radar signal based on sparse reconstruction parameter estimation problem. For three basic radar signal we improve modulation pattern recognition methods accordingly and estimate the three signals of typical parameters. Among them, for single carrier frequency signal, we focus on the estimate of the center frequency. Because the front method can estimate effectively, so we just improve algorithm, in order to improve the estimation precision. For LFM signal, we mainly aim at making an estimate of the frequency modulation slope. The fractional Fourier transform can cluster to the LFM effectively, so we can change sparse matrix's atoms to fractional Fourier transform of inverse kernel function. Changing order is obtained by finding the best clustering number, thus we can achieve the estimate of frequency modulation slope. For phase coded signal, the typical parameter is the element width and by calculating the inverse bandwidth the element width can be obtained. But due to the bandwidth for zero bandwidth, using of spare reconstruction error is too big, so we can only draw corresponding center frequency estimates. This area is still an open question.