On Robust Soft-Input Soft-Output Demodulators For Ofdm Systems: When Imperfect Channel State Information Is Present

In this paper a soft-input soft-output (SISO) QAM demodulator with robust performance on imperfect channel information are proposed for bit-interleaved OFDM systems. A full Bayesian approach is proposed to the channel estimation and demodulation problem. The frequency-selective fading channel impulse response and AWGN variance encountered in OFDM systems are jointly modeled as complex Gaussian-gamma random variables. The uncertainty of the channels are naturally encoded in the posterior distribution. Robust demodulators for known and unknown AWGN variance are derived basing on Bayesian posterior predictive distribution. It's performance combined with the bit-interleaved coded modulation (BICM) is demonstrated. And schemes with reduced complexity are also discussed. Simulation results show an improved BER (more than 0:5dB in most cases) comparing to that of the conventional demodulators ignorant of the channel estimation errors.