New fast parallel algorithm for adaptive multidimensional signal processing

Abstract

A technique to derive fast stable parallel adaptive signal processing algorithms is presented. The leading principle of the approach suggested is in using a fast-converged direct linear algebra method targeted to obtain an approximate basis of the solution decomposition, in combination with a parallel iterative tracking algorithm which uninterruptedly eliminates any deviations of the current basis from the optimal one. Based on the modified method of conjugate directions in combination with the adaptive parallel crossed feedback control technique a new parallel seamless algorithm has been developed for the antenna array adaptation problem. The adaptive algorithm shows a fast convergence rate: the number of iterations taken to achieve the effective interference suppression is proportional to the amount of undesired signals. In addition, the parallel algorithm has a number of peculiar properties to be effectively implemented in real-time signal processing systems.