Research Of Inverse Beam-forming Method Based On Fourth-order Cumulants

The research of DOA estimation in passive direction-finding sonar has important practice significance. The main problem is how to estimate the azimuth of targets when the number of targets and SNR are unknown. The usual spatial spectrum estimation algorithms always have some constrains. The article expects to design a tolerant spatial spectrum estimation algorithm which can be better used in the passive direction-finding sonar.The article first review the usual algorithms based on the second-order statistics such as conventional beam-forming, inverse beam-forming and subspace-like and analyze their performance. The theory analysis and simulation indicate: The azimuth resolution of IBF is improved 1/3 than that of CBF, and the array gain of IBF is improved 3dB than that of CBF. The azimuth estimation performance of subspace-like algorithms are much better than that of linear spatial spectrum estimation algorithms, but the problem of disparting the subspace is hard to solve when the number of targets is unknown, besides, the azimuth estimation performance is effected seriously by SNR and is worse than that of IBF in the low SNR circumstance. Therefore, IBF has better applicability if considering the tolerant performance.To improve the resolution of IBF further, the article proposed an IBF algorithm based on the fourth-order cumulants. High order cumulants has the ability of array expanding and restraining gauss noise. The usual subspace-like algorithms based on fourth-order cumulants improved the second-order subspace-like algorithms in a way, however the problem of eigenvalue choosing still limits it's apply. The article first get a minimum redundancy covariance matrix through Toeplitz average of fourth-order cumulants matrix, and then apply the technology of IBF, thus the length of array is expanded 2 times than former. Through using lowpass previous processing by moving average, the effect brought by high side-lobe is restrained without losing the information of power spectrum density and reducing the azimuth resolution. The simulation and history experiment data processing indicate: The azimuth resolution...