Signal processing algorithms based on Volterra filter structures often
require that a matrix composed of higher order moments of the underly
ing process be invertible. Previously, this problem has been studied f
or uncorrelated random variables. This paper establishes conditions un
der which a wide class of correlated processes have invertible higher
order moment matrices.