Linear adaptive decorrelator for signal separation

Signal separation is a problem often encountered in many practical applications. A novel technique is described that can guarantee signal separation by output decorrelation. In particular, it is shown that a simple transformation of the variables enables the nonlinear set of equations to be solved efficiently using the standard least squares technique. Moreover, the uniqueness of the solution is analytically determined and the system is shown to guarantee separation. The algorithm is modified to estimate the parameters adaptively, and the results clearly show the improved performance of this algorithm in separating two signals.