An improvement of the Lee and Schetzen cross-correlation method

Cross-correlation techniques have been used for years to identify the kernels of Volterra and Wiener nonlinear functional series. This note proposes a modification to the standard cross-correlation method which improves the estimation of the diagonal elements of the kernels. The mathematical derivation is given and simulations are presented, which show that modification improves the kernel estimates with very little increase in computational cost.