Adaptive Subspace Estimation With The Application In DOAs Tracking

Multiple source DOAs (directions of arrival) estimating and tracking technique plays a very important role in sonar, radar, communication, national defense, e.t.c. Aiming at the applications in engineering, the principles of subspace algorithms about the DOAs estimating and tracking are discussed comprehensively in this dissertation. Considering the characteristics of underwater detecting and tracking system, the subspace algorithms about multiple source DOAs estimating and tracking are systematically studied by theoretical analysis and computer simulation. The main results and achievements are summarized as follows:1. The subspace high-resolution direction-finding methods are disserted, and the formulas for the MUSIC, Mini-Norm, Root-Music, and ESPRIT methods and the calculation procedures are discussed also. The integrated performance evaluation indexes are presented. According to these indexes, a great deal of computer simulation is done and the performances of the aforementioned subspace methods are analyzed. Based on the analysis, the practicability of the methods is evaluated.2. The adaptive subspace estimation algoritms arc investigated thoroughly and can be broadly categorized as follows: one is modified eigen problem (MEP) methods, the other is non-MEP or adaptive methods. In MEP methods, the classical EVD algorithms are extended from the stationary case to the non-stationary case via the rank-1 or rank-2 modified data covariance matrix. In non-MEP methods, the exact eigen information is not computed at each update, and the results can only move towards to the EVD (or some aspect of the FiVD) information. On the basis of these, two adaptive subspace estimation algorithms are introduced and studied. One is the MALASE algorithm proposed by Chonavl et al. which belongs to the non-MEP methods, and the orthononnality of the estimated basis of eigenvectors of MALASE can be ensured because of its particular structure. Its computational complexity is O(M~2)or O(Mr). The other is the ACN algorithm proposed by Babu et al, which belongs to the MEP methods and uses a pipeline architecture, so it is highly modular and parallel, and requires O(M~2) operations.3. Three adaptive high-resolution algorithms are proposed. One is the adaptive Root-MUSIC algorithm based on MALASE. one is the adaptive Mini-Norm algorithm based on MALASE. and the third is the adaptive Root-MUSIC algorithm based on ACN. The three algorithms can all track the moving DOAs for real applications on a snapshot basis. Computer simulation results are provided to demonstrate the ideal DOA-tracking effectiveness of the proposed three algorithms.