Abstract
A novel class of binary parallel concatenated recursive systematic convolutional codes termed turbo-codes, having amazing error correcting capabilities has been proposed recently. Several recent studies have demonstrated that turbo codes can also achieve remarkable bit error performance over flat Rayleigh fading channels. When used in a multipath fading channel, modification of the decoder is needed. Hall and Wilson have investigated the performance of turbo codes in either independent or exponentially correlated slow Rayleigh fading, assuming perfect channel estimation is available. In pratical systems, the channel must be estimated at the receiver. Several estimation methods, such as lowpass filter or pilot symbol assisted modulation technique, have been used to perform channel estimation prior to turbo decoding. In this thesis, we investigate the LMS and the GOBA adaptive filter to estimate the channel information by taking the absolute value of the received signals as the input and the desired signals. We have recognized that the absolute value operation will limit the capability of the channel estimator when suffering deep fades. We have also proposed a new structure that integrates the estimation process and the turbo decoder into iterative manner. In such a way, the results of the decoder during each system iteration are used to assist the estimation for the next iteration. The performance of the proposed structure have been simulated by using the LMS adaptive algorithm and the GOBA algorithm, respectively. Computer simulation results have indicated that the GOBA channel estimator exhibits excellent bit error performance which is close to the ideal case (with perfect knowledge of channel state information) for BdTs = 0.01.