Pilot symbol assisted QAM with iterative filtering and turbo decoding over Rayleigh flat-fading channels

We investigate the application of iterative channel estimation and decoding for turbo coded M-QAM signals transmitted over slow Rayleigh fading channels. Pilot symbols are inserted periodically in the encoded data stream for the estimation of the time-variant channel state information (CSI). The decoder uses this estimated CSI, together with the received data sequence, to generate soft decisions for each data symbol. These soft decisions and the known pilot symbols are then filtered again to obtain improved estimation of the CSI. Two different algorithms, algorithm A and algorithm B, have been studied for the iterative filtering process and their performances are evaluated by simulation. At a BER of 10/sup -3/, both algorithms are within 1.3 dB of the ideal system with perfect CSI. Compared to the exponentially growing complexity of the optimal joint channel estimation and decoding system, the proposed system has a linear complexity when algorithm A is used and a polynomial complexity when algorithm B is used, thus providing low-complexity alternatives while maintaining an effective performance improvement.