|
Abstract:
|
This thesis has been done in the Signal Processing Laboratory at Tampere University of Technology in collaboration with Nokia Mobile Phones. This project was a part of research program called "Adaptive and Intelligent Systems Applications", which was set up by Technology Development Centre (TEKES). Equalization is one the most important tasks of a digital communication receiver. The purpose of equalization is to reduce the effects of communication channel to the transmitted signal in order to be able to found out the original information being sent. In this thesis we have first discussed conventional equalization methods, and introduced three different adaptive equalization techniques. The introduced methods are adaptive linear equalizer, MLP neural network (Multilayer Perception) and adaptive clustering method. These methods have been used in equalization of binary data bursts, that have a structure, which reminds the one used in GSM system. These three methods have been used for equalization in both fixed and altering channel. The comparison of the methods is done by observing the bit error rates achieved with each method. In addition the computational complexity of each method has been observed. The best bit error rates were achieved with adaptive clustering and MLP network. The computational load of adaptive clustering and linear equalizer was smallest. /Kir10 |