Free-space optical communication has been an area of intense research for many years, primarily due to the greatly increased data bandwidths available, compared to RF systems. However, atmospheric effects cause free-space optical communication channels to have a highly variable transmission. Therefore, due to multiplicative, or signal-dependent, noise in the receiver detectors, the optimum decision threshold for bit detection varies with time. The work described in this dissertation concerns development of methods for prediction of communication bit parameters to allow adaptive updating of the detection threshold. I have developed four adaptive thresholding techniques, two based on Kalman filter predictors and two based on adaptive linear transversal filters (Least-Mean-Squares and modified Sequential Regression) to track changes in bit level mean and variance values and to adaptively update the detection threshold to maintain near optimum. These methods are shown to improve the obtainable bit error rates to the theoretical limit and are more than an order of magnitude improvement over non-adaptive thresholding methods. |