In this paper we present a new method for the efficient implementation
of the fast transversal filter (FTF) algorithm. Reduction of the arit
hmetic complexity is obtained by making use of the redundancy in the s
uccessive computations of the forward prediction error and the filteri
ng error in the joint process. The resulting algorithm is exactly equi
valent to the original FTF algorithm, hence retaining the same theoret
ical convergence characteristics and offering the least squares (LS) e
stimate at each recursion step without delay. Furthermore, the algorit
hm can be numerically stabilized by using a simple and effective stabi
lization measure which needs only one additional multiplication per re
cursion step. The equivalence of the proposed algorithm to the origina
l FTF algorithm is demonstrated by simulations of an acoustic room imp
ulse response identification.