A method is discussed to separate blind source signals only based on second order statistics.The first step is to estimate the relation matrix of the sensor signals

getting main eigenvalues and the corresponding eigenvectors of signals through eigenvalue decomposition.Simultaneity

the singular values of the pre-whited time-delayed signals can be obtained by way of singular value decomposition.Applying these informations

the separated signals can be estimated.The simulations show that the algorithm has low computing cost and good separation performance to the blind mixed signals.