Reduced Complexity Adaptive Demodulation in Impulse Noise Channels

Modern vehicles, military and civilian, often contain several closely located electronic systems that create electromagnetic interference. Such interference is often highly non-Gaussian and a more suitable statistical model than the Gaussian model is necessary to derive well-functioning receiver algorithms and to analyse the communication performance. In this work, we consider the more general Symmetric α-Stable (SαS) distribution. Demodulation is performed by computation of a log-likelihood ratio, which for the for SáS model boils down to computationally burdensome numerical solutions. Therefore, we propose to reduce the overall computational power of an interference adaptive demodulator by determination of when the noise characteristics are such that using the SαS model would provide a significant gain and when the standard Gaussian model performs well enough. The proposed algorithm is based on estimation of the SαS parameter α and $E_{b}/N_{o}$ which are compared to predetermined thresholds for the desired BEP to determine which demodulator to use. Numerical results show that the gain of the proposed algorithm, in terms of the $E_{b}/N_{o}$ required to meet a desired BEP, is in the order of 20 dB compared to using the standard Gaussian demodulator. Moreover, the performance is similar to using the optimal SαS demodulator but with significantly reduced computational power on the average.