Serial Greedy SMLR Impulsive Noise Detection and Channel Equalization for OFDM Systems

Impulsive noise may be generated by vehicle ignitions, electric device switching, or strong bursty radio frequency emissions. It will severely degrade orthogonal frequency division multiplexing (OFDM) based wireless vehicular communication systems. In this paper, the single most likely replacement (SMLR) algorithm with the serial greedy iteration is proposed to detect the Bernoulli-Gaussian impulsive noise for OFDM based wireless systems. Compared to the conventional SMLR algorithm, the serial greedy SMLR algorithm updates multiple elements per iteration, significantly reducing the required iterations. Moreover, by employing approximate substitution and matrix partitioning, the proposed algorithm reduces its computational complexity from cubic in the discrete Fourier transform dimension to nearly linear, while maintaining detection performance comparable to the conventional SMLR algorithm. Based on the algorithm's detection results, an impulsive noise aware channel equalization is proposed to simultaneously equalize the channel and suppress impulsive noise. Formulated under the minimum mean square error (MMSE) criterion, the method accounts for the correlation introduced by multipath channels and does not rely on null subcarriers for noise suppression. Simulation results show that the proposed impulsive noise aware equalization can achieve a great performance gain over other suppression methods when the impulsive noise is not highly sparse in the time domain. For example, at a target bit error rate (BER) of 10^-4, it achieves about a 3.5 dB gain over sparse Bayesian learning when the impulsive noise occurrence probability is 0.1 and the impulsive noise-to-background noise ratio is 20 dB.