Covariance matrix representation problem arised especially from the necessity of the knowledge of covariance matrix information in track fusion algorithms. "Representation" means representing the covariance matrix with its exact or approximate parameters. In this study, representation methods for covariances matrices are proposed and approximate representations is analyzed by Kullback-Leibler distance. We assume that state vector is Gausssan distributed and the means are the same for both the representation and the actual states. Performance of the representations is measured by generating synthetic tracks and quantizing the representation covariances.