The Performance Analysis Of Spatial Spectrum Estimation Algorithms

Spatial spectrum estimation algorithms detect signals in space and locate their positions through space array, its main objective is to improve the accuracy, the angle resolution and the operation speed with no loss of performance. Array super resolution algorithms can break Rayleigh limit in angle estimation, so show a broad application in communications, radar and other areas with excellent estimation performance.MUSIC algorithm is the classic representative of super resolution algorithms of subspace, which is a milestone in spatial spectrum estimation research. Firstly, the classical MUSIC algorithm, ESPRIT algorithm and beam space MUSIC algorithm are briefly introduced in this paper, then simulates and compares their performances. MUSIC algorithm has a high resolution, estimation accuracy and stability in the condition of large signal to noise ratio(SNR) and a large number of snapshots, in the ideal case, the performance is close to Cramer-Rao bound. But the disadvantages of the subspace super resolution algorithms are that they need to estimate the covariance matrix and eigen decomposition of covariance matrix, the computation is so large that their application in engineering is limited.In order to avoid the eigen decomposition of the covariance matrix, this paper studied the MUSIC algorithm with no eigen decomposition of the covariance matrix, The MUSIC algorithms based on data conjugate rearrangement, difference spatial smoothing, Wiener filter structure and Schmidt orthogonal (which is proposed in this paper) are introduced. Then simulates these algorithms and compares their performances, simulation results show that the performances of those algorithm are close to the MUSIC algorithm when SNR is large enough (15dB above).To avoid a large number of snapshots in MUSIC algorithm, a single snapshot MUSIC algorithms are studied. A single snapshot can use less data, therefore, SNR, noise and source number are highly required. MUSIC algorithms based on dimension reduction, Hankel matrix, Wiener filter structure are focused on study in this paper. These algorithms are all needed to reduce the dimension of the array, which reduce the array's effective aperture and the freedom of the antenna. Finally, these algorithms are simulated, the results show that performances of the algorithms are declined compared with the large snapshots of the MUSIC algorithm in a low SNR.Signal processing feature of high frequency ground wave radar is that detect target peak area determination after range-doppler processing, then super resolution estimation of angle is applied in a particular rang-doppler unit. This paper deduces the array model of a single snapshot DOA estimation in frequency domain in high frequency radar. simulation results show that after dimension reduction the orthogonality between noise subspace and signal direction vector is better than before, so verify the necessity of the dimension reduction in a single snapshot. Against the low resolution of the MUSIC algorithm based on dimension reduction, finally, this paper proposes a new MUSIC algorithm based on dimension reduction after signal subspace projection, both simulated and the real data results verify the availability of the new algorithm.