New forms of RLS ladder algorithms for array processing

Abstract

Summary form only given. Two recently developed triangular array ladder algorithms are discussed. The first algorithm, ARRAYLAD 1, computes both the transversal forward/backward predictor coefficients, the ladder reflection coefficients, and the forward/backward residual energies. This is obtained at a total computational complexity of 1.5 p/sup 2/ multiplications per recursion, where p is the order of the algorithm. ARRAYLAD 1 can therefore be implemented on a triangular systolic array with three multipliers per rotational (triangular array) element in the scalar (single-channel) case. The second algorithm, ARRAYLAD 2, computes only the ladder reflection coefficients and the residual energies at a reduced computation complexity of 1.0 p/sup 2/ multiplications per recursion, hence requiring only two multipliers per triangular array element in a systolic array implementation.<>