A system and method in a radio receiver for joint synchronization and noise covariance estimation of a received signal. A spatially and temporally stacked signal model, whereby successive samples of temporally adjacent received signal vectors and corresponding training vectors are stacked, is used in the derivation of the estimation problem. The Toeplitz structure of the channel response matrix is neglected in the formulation of the estimation problem. The resulting estimator jointly estimates a synchronization position, a channel response matrix, and a noise covariance matrix. An estimate of a whitened channel is then computed based on the noise covariance matrix and the estimate of the channel response matrix.