High Resolution Spectrum Estimate Research With Application In Radar Signal Processing

The radar imaging technology is one of the most important aspects of modern radar technology which has been applied widely in both military and civil areas. It is of great importance in the precision guidance application. The high resolution range profile (HRRP) and the SAR two-dimensional resolution images contain rich information such as the target's structure and distribution etc, therefore, they own sound prospects in target detection, target recognition, and high accuracy tracking.According to multi-scattering centers model, radar's echo wave can be described with a one-dimensional or two-dimensional harmonic model. Therefore, the issue of radar imaging is equivalent to that of spectrum estimation.This thesis points out the advantages and disadvantages of the classical and modern spectral estimation methods respectively through analyzing and stimulating. The results reveal that the classical spectrum estimation has low frequency resolution, but its calculation is simple and amplitude information can be restrained; while super-resolution spectrum estimation has high resolution, however, it has the disadvantages of complex calculation, the loss of amplitude information based on the pseudo-spectral function and high sensitivity to noise. Because of these shortcomings, two improved methods of spectrum estimation are proposed in this thesis: optimal window surface integral spectrum estimation based on fourth-odder cumulant and spectrum estimation based on the noise power compensation and the limited spectrum least squares.Two new researches of this thesis are mainly as follows:(1) In the condition of small samples, the conventional method of fourth-odder cumulant diagonal slice has the problem of signal model distortion and the performance of spectrum estimation is weakened. Therefore, a higher-order statistic of harmonic which can be used in small samples'spectrum estimation is designed: the optimal window surface integral spectrum estimation based on forth-order cumulant and the harmonic spectrum estimation model based on this statistics is also proposed. Theoretical analysis and simulation results show that this improved method has a better resolution performance and parameter estimation precision compared with traditional forth-order cumulant.(2) Considering the loss of amplitude information and complexity of calculations which are the main problems of traditional super-resolution spectrum estimation, spectrum estimation based on the noise power compensation and the limited spectrum least squares estimation is proposed. Theoretical analysis and simulation results show that this improved method reserve super-resolution performance, significantly reduce the operation spending and has a high precision of amplitude parameter estimation.