Research On Blind Adaptive Algorithms Applied To Antenna Arrays

Nowadays, with the further enhancement of the capacity and reliability of the communica-tion systems, how to improve the algorithms for antenna arrays and reduce their complexity, has become a problem of growing concern. As with the direction-of-arrival (DOA) estimation issue in antenna array systems, the complexity and performance of the DOA estimation algorithms be-come key factors of the availability of large-scale antenna arrays. In the beamforming issue of the antenna arrays, the adaptive beamformer need to resist the fast time variability and instability of the communication system. This thesis study the above problems in-depth and propose effective improvement algorithms.In this thesis, an alternating low-rank decomposition (ALRD) approach for DOA estimation in a large-scale antenna array system is proposed. In the ALRD scheme, a subspace decomposition matrix which consists of a set of basis vectors and an auxiliary low-rank parameter vector are employed to compute the output power spectrum for each scanning angle. In order to avoid matrix inversions, we develop recursive least squares (RLS) type algorithms to compute the basis vectors and the auxiliary parameter vector, which reduces the computational complexity. The proposed DOA estimation algorithms are referred to as ALRD-RLS and modified ALRD-RLS (MALRD-RLS), which employs a single basis vector. Simulations results show that the proposed ALRD-RLS and MALRD-RLS algorithms achieve superior performance to existing techniques for large arrays with short data records.As with the design of adaptive beamformers, we derive and simulate the combination of two structures and two adaptive algorithms based on the linearly constrained minimum variance (L-CMV) criterion and choose the combination with the best performance, i.e., the beamformer with the generalized sidelobe canceller (GSC) structure and the RLS algorithm. Then two variable for-getting factor (VFF) mechanisms are proposed. The first proposed mechanism employs the time-averaged cost function to update the forgetting factor, which is referred to as the time-averaged variable forgetting factor (TAVFF) mechanism. The second proposed scheme named the correlated time-averaged variable forgetting factor (CTAVFF) mechanism applies the time-averaged correla-tion of the cost function to the update equation of the forgetting factor. The steady-state analyses of the LCMV-RLS-GSC beamforming algorithms with the proposed VFF mechanisms are carried out in terms of the computational complexity and the convergence properties. We also investigate the effects of the parameters on the performance of the proposed mechanisms. Simulation results show that the LCMV-RLS-GSC beamforming algorithms with the proposed VFF mechanisms achieve superior performance with a significantly reduced complexity.