Adaptive algorithms for channel estimation: Using a priori information for optimal design.

摘要

This thesis is concerned with using a priori information on the communication problem to enhance the performance of adaptive algorithms. We demonstrate this for two scenarios.;In the first part of the thesis, we consider the problem of channel estimation for multiuser OFDM transmission over block time varying channels. We show that frequency domain based channel estimation can be more practical and less computationally complex than the time domain based estimation. We show how a priori information (time and frequency correlation) allows us to design Kalman filters and obtain a very good performance in environments with high Doppler.;In the second part we use our a priori knowledge about the output noise distribution to design adaptive filters with optimum error nonlinearities. Specifically, we use the energy relation that is usually used for mean square analysis to derive the mean square error for general error nonlinearities and subsequently use that to design optimum error nonlinearities given our knowledge about the noise distribution. We use a similar technique to design adaptive filters with optimum error nonlinearities and data normalization that is usually used when the input data is correlated.