The present disclosure relates to an iterative method for estimating covariance matrices of communication signals comprising a) computing a reference symbol covariance matrix estimate (k,l); b) inputting said reference symbol covariance matrix estimate (k,l) to a detector or a decoder in a first iteration, and thereafter inputting the covariance matrix estimate output from e) to the detector or the decoder in subsequent iterations to obtain an updated demodulated or decoded communication signal for each iteration; c) inputting the demodulated or decoded communication signal to a symbol generator; d) computing an updated data covariance matrix estimate (k,l) for each iteration based on data symbols of said regenerated communication signal; e) combining said reference symbol covariance matrix estimate (k,l) and said updated data covariance matrix estimate (k,l); and f) forwarding said covariance matrix estimate output to the detector or decoder in b) to obtain the updated demodulated or decoded communication signal for each iteration.