This paper presents a new incremental algorithm for the computation of characteristic (weighting) patterns and vectors of discrete convolution systems. The method first approximates the system by a real characteristic dyadic form. Considerable savings in computation time are made if this is a good approximation. If not, the estimates are incremented at each instant to find the actual characteristic patterns and dynamic vectors. Numerical examples are given for real dyadic and non-dyadic forms.