Abstract: A method generates signals in a transmitter of a multiple-input, multiple-output wireless communications system. The transmitter includes Nt transmit antennas. A transmit covariance matrix Rt determined using statistical state information of a channel. The transmit covariance Rt matrix is decomposed using transmit eigenvalues ?t to obtain a transmit eigenspace Ut according to Rt=Ut?tU†t, where † is a Hermitian transpose. A pilot eigenspace Up is set equal to the transmit eigenspace Ut. A Nt×Tp block of pilot symbols Xp is generated from the pilot eigenspace Up and pilot eigenvalue ?p according to Xp=Up?p1/2. A data eigenspace Ud is set equal to the transmit eigenspace Ut. In addition, a Nt×Nt data covariance matrix Qd is generated according to Ud?dU†d, where ?d are data eigenvalues. A Nt×Td block of data symbols is generated, such that an average covariance of each of the columns in the block of data symbols Xd equals the data covariance matrix Qd.