Abstract
One of the most well-known adaptive algorithms that adjust conveniently the parameters of the TV ghost canceler is the least mean square (LMS) algorithm. The convergence rate of this algorithm is slow. In this paper a parallel processing technique to speed up the computation of the least square (LS) algorithm is studied. The technique is based on mapping the recursive modified Gram-Schmidt algorithm into two levels of linear systolic arrays. The arrays structure as well as the computation time are described. The performance of the system in terms of computational saving and the array utilization are evaluated. The system is shown to be composed of identical modules which makes it suitable for VLSI technology implementation. The great saving in computation will qualify the system for real-time applications.