Intersymbol Interference (ISI) is caused by multipath effects or bandwith constraints and presents a major obstacle in modern high- speed digital transmission. When the communication link is also fading with time, it is essential for the equalizer to be able to follow the channel's time variations. While several adaptive and blind methods have been proposed for time-invariant or slowly fading channels, relatively little attention has been given to the rapidly fading case. The major goal of this project is to use basis expansion ideas as a tool for the equalization of rapidly varying channels. In this framework, each time-varying coefficient is expanded into a set of basis sequences, and the expansion parameters are estimated using novel adaptive and decision directed schemes. Next, basis expansions are combined with time-varying correlations and higher- order statistics to address time-varying, blind equalization in a rapidly fading environment. Finally, based on physical intuition and experimental data, the selection of the "best" basis which captures optimally the channel's dynamics is being addressed. Complex exponentials are being investigated for the mobile radio channel, while wavelet expansions are being considered for channels with abrupt changes, and are being compared with stochastic channel modeling approaches.
