Research On Denoising And Parameter Estimation Of LPI Radar Signal

Detecting and processing LPI (Low probability of intercept) radar signals correctly is the key of the research, because the appearance of LPI radar brings a new challenge to the anti-radiation missile seeker. The application of signal denois-ing and parameter estimation to passive radar seeker was investigated in this dissertation, whose major research issues were how to filter the noise quickly and how to estimate parameter correctly at low SNR.In normal variable step size LMS (Least Mean Square) algorithm, ideal signal was taken as reference signal, so as to get the convergent error signal to control the step size. However, step size was not zero when the ALE (Adaptive Line Enhancement) system reached steady state, which made the iteration continue, because the ideal error was approximate to broadband noise. Then a novel variable step size LMS algorithm applied in ALE system was presented. It built a nonlinear relationship between the step size and the weighted coefficient, which made step size decrease till zero with weighted coefficients decreasing. In order to accelerate the convergence further, it also introduced the step-vector to adjust each weighted vector value real-timely. And then it took the LFM signal and NLFM signal as examples to verify the validity of algorithm.Adaptive denoising method based on DWT (Discrete Wavelet Transform) provided a feasible solution for radar signal filtering. However, DWT doesn't have the characteristic of translation invariance, so it would bring bigger reconstruction error and impact the filtering effect if different wavelets are used to reconstruct signals. The lifting methods of SWT (static wavelet transform) and static wavelet packet transform were introduced, and then the adaptive denoising methods based on SWT and static wavelet packet transform was proposed. It decomposed the signal into multi sub-bands by lifting SWT and static wavelet packet transform, and matched adaptively using iterative formula with weighted coefficients which had more momentum factors. Finally, it took the second adaptive filter of the matched results to acquire the fitting signal. Both of the new methods can enhance SNR of the output, whose calculation amount is a little more than traditional method.The algorithms represented by HAF (high-order ambiguity function) and PHAF (product of the high-order ambiguity function) are sub-optimal methods to estimate parameters of PPS (Polynomial Phase Signal). But these methods have the disadvantages of existing SNR loss, leading to accumulated error, causing missed detection and false detection. A RFRFT (Reduced Fractional Fourier Transform) algorithm was introduced by eliminating computational redundancy of FRFT algorithm. Meanwhile, the expression of RFRFT modulus square of LFM signal was given, the relationship of estimated parameters before and after the normalization was discussed. And resolution of RFRFT was enhanced by using angle transformation. Afterwards, a PRFRFT (Product Reduced Fractional Fourier Transform) algorithm was proposed according to phase difference method and the different characteristics of auto-terms and cross-terms of mc-PPS in RFRFT. It accomplishes parameter estimation of mc-PPS with frequency above MHZ. And it has low computational complexity and easy implementation. Simulation results prove that this method is able to suppress the noise and cross-terms in low SNR.Signal parameter estimation based on cycle correlation theory is popular because of no need of priori information, but its estimating effect of code rate is not satisfactory in low SNR. Moreover, it has a large amount of computation due to the two-dimensional search. Firstly, a parameter estimation algorithm based on improved cycle auto-correlation was proposed. Considering that the characteristics of cycle auto-correlation of PSK (phase-shift keying) signal was decided by its carrier frequency and code rate, two statistics based on the cycle auto-correlation function were established to achieve blind parameter estimation. Secondly, a fast PSK signal parameter estimation algorithm was proposed. It recognized the PSK signal, considering that the amplitudes of cumulative outputs of baseband signal's imaginary parts and real parts were complementary and their inflexion points were just the jump point of the phase. This method is easy to calculate and implement. We can select one of the methods appropriately to estimate parameters of PSK signals in practice.