Channel Estimation in Massive MIMO with Heavy-Tailed Noise: Gaussian-Mixture Versus Cauchy Models

Impulsive noise can appear in communication links. In the literature, it was demonstrated that when the noise is impulsive, standard Gaussian receivers perform poorly because of the outliers in the noise. Therefore, appropriate receivers must be used when the noise is impulsive. In this paper, we compare two types of massive multiple- input multiple-output (MIMO) receivers, namely those based on a Gaussian-mixture assumption and those based on a Cauchy assumption, in terms of channel estimation quality, when the noise is impulsive. Symmetric α-stable (SαS) noises are used to model impulsive noises in the paper. In the numerical results, the Gaussian-mixture receiver outperforms the Cauchy-based receiver.