Abstract
Multipath effects occurred in digital communications may degrade the performance enormously. In order to eliminate multipath effects, adaptive equalizers are frequently used in the receiver. The conventional adaptive equalization approach includes a training mode equalization algorithm (LMS, RLS, GOBA) which is used to estimate the equalizer coefficients as accurately as possible. However, while training sequences are not available or expected, blind equalization is necessary. The recent blind method proposed by Gardner and Tong based only on second-order statistics (SOS) was a major breakthrough. The basic idea of this new approach is to take advantage of additional time or spatial diversity at the receiver. This additional diversity transforms that a single-input single-output (SISO) channel estimation problem into a single-input multiple-output (SIMO) problem. In this thesis, we devote to design an adaptive blind equalization approach which is robust to channel order overdetermination and has lower computational complexity. We further extend our scheme to the multiple delay case. This approach enables us to estimate any non-zero delay data. The advantage of this approach is that if the zero-delay data can be estimated accurately enough when is small, it is possible to estimate the non-zero delay data more accurately.