Fundamental Studies On Extracting Parameters Of Frequency Agile Signal

Signal parameters extraction technology is mostly applied to the estimation of the signal wave and parameters from the background noise. The useful object information that could be detected includes target numbers, frequency, amplitude, bandwidth and phase, etc.Frequency Agile Signal (FAS) is a kind of non-stationary signal, as the result of the high-speed, random agility of the carrier wave frequency of two neighbor emissive pulses or pulse arrays within the certain range. In this dissertation, some fundamental researches have been done on FAS parameters extraction technology, mainly on parameters extraction, including components of signal, etc. by time-frequency analysis. The paper is consisting of two parts: time-frequency analysis about FAS and high resolution instantaneous frequency (IF) estimation for FAS. With this research we hope to get sufficient knowledge of FAS processing and strengthen FA signal anti-jamming capability to achieve the best underwater acoustic counter measures effect.The task of time-frequency analysis is to describe how signal spectrum clearly changes with time and presents signal IF on time-frequency plane. Based on it, analysis and comparison has been made among three kinds of commonly-used time-frequency distribution (TFD) from theory and simulation. Simulation result indicates that, Spectrum is advantaged in simple calculation and precise definitions, so it can be regarded as a practical TFD. For linear frequency modulated (LFM) signal, Wigner distribution (WD) seems nearly to be the best TFD, which can attain the high concentration along the IF. Polynomial Wigner distribution (PWD) offsets the disadvantage of WD in that it generates artifacts when processing none-linear modulated signal or multi-component signal. PWD can get FM signal IF direction of arbitrary orders, also fits for multi-component signal. Spectrogram-Wigner distribution (SWD) is proposed by paper. SWD is a transform between a spectrogram and WD. Moreover, it will combine the good properties of both. It uses the time-finite window of spectrogram to avoid cross terms. It makes up the significant leakage of spectrogram due to the window selection and inherits higher resolution of time-frequency from WD. This chapter puts emphasis on B-distribution. How to design the kernel function of B-distribution and how to discrete it, also its characters have been discussed in detail. The B-distribution based on Radon transform (BDBRT) is proposed by paper. Based on that Radon can be integral along the line, moreover, IF of LFM and CW is a line on time-frequency plan, Radon transform and B-distribution are conjoined together. It can recognize two signals which are very close and does not suffer from cross terms. STFT and WD can not reach it. Results on the trial data show that the arithmetic is more useful than others.IF is an important parameter of FAS. This chapter puts more effort into the IF estimation based on TFD. At the start of this chapter, Wigner distribution estimator is discussed. And we deduces the LFM signal IF estimation Cramer-Rao Lower Bounder (CRLB). The dissertation also proves that the theory estimated variance of no interpolated WD estimator equals to CRLB of LFM signal; but interpolated WD estimator's increases 3 times higher than the former, because of the effect of interpolated factor. Simulation presents when signal-to-noise rate (SNR) higher 5dB, no interpolated WD estimator is unbiased for LFM signal. Then we discuss the high-order TFD estimator. By using the peaks of PWD as IF estimator, we can get the theory variance of quadratic FM signal. The results indicates when SNR is higher 5dB, the variances of six-order PWD can reach the quadratic FM signal's CRLB. Afterwards, B distribution estimator is focused on. We deduce its IF theory variance. Comparison has been made between B distribution and Choi-Williams distribution by simulation. It proves that B distribution lobe near the signal IF is narrower than Choi-Williams'. And the variance value, the former is lower than the latter. A comparison between B distribution and WD has been made in analysis and IF estimation of LFM signal. At last, trial data also are processed. The results prove both B distribution estimator and SWD estimator can estimate FA signal's frequency precisely and they are high accuracy IF estimators.