Research On Fast MUSIC Algorithm Based On UCA

The MUSIC algorithm based on uniform circular array(UCA) has excellent performance,but its huge computational complexity has been limiting its use in occasions with high real-time requirements. The attention in this paper is mainly concentrated on the fast MUSIC algorithm based on UCA, the details are as follows:(1) UCA-RB-MUSIC algorithm first employs phase mode excitation-based beamforming to synthesize a beamspace manifold similar to that of a ULA, and then adopts unitary transformation to make beam-space data covariance matrix real-valued. This paper proposes a unitary transformation matrix different from UCA-RB-MUSIC algorithm, which reduces the computational complexity without sacrificing performance.(2) Subspace estimation occupys a pivotal position in the high-resolution DOA estimation, but it is generally necessary to calculate the data covariance matrix and its eigenvalue decomposition,which needs a huge amount of computation time especially when the number of array elements is large. As for this problem, this paper first has a deep study of the eigenvalue decomposition algorithm for hardware implementation, and then proposed an improved one-sided Jacobi algorithm,which uses a small hardware overhead to achieve automatically sorting of eigenvectors as eigenvalues size during the iteration. Secondly,subspace estimation based on multistage Wiener filter (MSWF) has been studied in this paper.Meanwhile this paper gives an improved method for its sharp decline performance with low SNR. This method improves subspace estimation performance based on MSWF under low SNR, but it at the same time brings a problem that there is a certain probability of spurious peaks.(3) While researching on DOA estimation, the number of sources is generally assumed known, but it is need to be estimated in practical application. For this, the estimation of the number of sources based on MSWF has been studied in this paper, and an improved method for its sharp decline estimation performance in low SNR is proposed.(4) Subspace estimation in UCA-Root-MUSIC usually has a large computational complexity. As for this problem, this paper uses subspace estimation techniques based on MSWF. Meanwhile, the paper uses a preprocessing method in MSWF of UCA-RB-MUSIC algorithm and its improved algorithm,which transforms the complex operation of MSWF into efficient real operation. Subspace estimation based on MSWF in UCA still has sharp decline performance in low SNR. As for this problem, this paper proposes another improved method.The method greatly reduces effection of noise in the performance of subspace estimation according to the essential reason for performance degradation, but this method requires a pre-estimated noise power. Subspace estimation based on MSWF has a variety of combinations with UCA-Root-MUSIC, the paper makes a large number of simulation experiments to analysis their performance.