In a usual MIMO framework when noise is assumed Gaussian, the channel matrix can be estimated with a Least Square approach. In addition, the Euclidean distance is optimal to detect the transmitted symbol. However, in dynamic interference, the Gaussian assumption is no longer valid. In this case, the alpha-stable model is better suited and the receiving strategy fails. In this paper we present a MIMO communication system based on QPSK modulation, Alamouti coding, 2 transmitting antennas, M receiving antennas and supervised channel estimation. We propose to modify both the channel estimation method and the distance metric to improve the receiver strategies. The Least Absolute Deviation estimator is used to estimate the channel matrix, instead of Least Squares, and the p-norm replaces the Euclidean distance (for which p = 2), with an estimator for the value of p. Our approach outperforms the traditional method.