Thèse de doctorat en Sciences pour l'ingénieur
Sous la direction de Samir Saoudi.
Soutenue en 2006
à Télécom Bretagne .
Transmission of broadband signals over a multi-antenna multipath fading channel induces intersymbol interference (ISI) at the receiver end. A main task of the space-time signal processor in reception is to combat the ISI effect with the aid of channel equalization techniques. An attractive alternative in case of coded transmission, is turbo equalization which consists on iteratively exchanging decisions between the equalizer and the channel decoder in the form of log-likelihood ratio (LLR) values, to improve information bit decisions at each iteration. This thesis focuses on efficient turbo equalization and multiuser detection schemes for coded space-time transmission over MIMO-ISI channels. We propose a low-complexity turbo receiver architecture for performing channel estimation and equalization iteratively with channel decoding for single user coded space-time transmission, and provide its variant in case of unknown co-channel interference (UCCI)-limited MIMO channels. The soft equalization concept relies on a successive interference cancellation (SIC) scheme derived under a minimum mean square error (MMSE) formalism. It particularly allows low computational complexity compared with existing techniques. The channel estimation algorithm is also MMSE based, and exchanges soft information with the channel decoder and the soft equalizer to improve channel state information (CSI) reliability at each iteration. A second contribution of this thesis is a reduced complexity iterative receiver technique for multiuser multi-antenna coded systems. The considered framework supposes that users employ different pilot sequences. At the receiver end, multiuser interference (MUI) cancellation, channel estimation, equalization, and channel decoding are iterated to reduce each user's error rate at each iteration. Efficient covariance matrices estimation tools are employed for dramatically decreasing receiver's computational complexity. Simulation results show that with a number of receive antennas less than the total number of transmit antennas in the system, the proposed scheme allows good bit error rate (BER) performance compared with the single user bound.
Pas de résumé disponible.