Improved blind adaptive equalization algorithms and analysis

Blind source separation (BSS) techniques allow recovery of individual unobserved signals from their observed mixtures, exploiting only the assumption of mutual independence of sources. Blind equalization combats signal distortion in digital transmission without knowing the channel response or the transmitted signal. In this dissertation, several new improved blind adaptive equalization algorithms are proposed, which exhibit better performance than their conventional counterparts. The convergence behavior of some of these algorithms is also investigated through analysis and simulation. In addition, some BSS approaches are proposed for source separation and phase recovery.; By using independent in-phase and quadrature components of a signal constellation, BSS techniques can be designed to estimate the carrier phase of a single signal. We propose a sector rotation scheme used in conjunction with a BSS algorithm for phase tracking of PSK signals. For multiple mixed sources, sources can be separated with phase recovery without affecting separation performance.; For conventional blind equalization which relies only on the overall statistical characteristics of the signal constellation, adding more detailed knowledge about the constellation points can help to improve equalizer performance. We propose several modified algorithms that add a constellation-match error (CME) function to the standard cost functions for blind equalization. The CME function provides detailed information of signaling constellations. Dynamic convergence behavior of some of these algorithms is also analyzed. It is observed that these analysis results agree very closely with the corresponding simulation results. Both simulation and analysis show the good performance of the proposed algorithms, which converge faster and/or have lower residual errors than the conventional blind equalization algorithms.