Efficient implementation of class of isotropic quadratic by using Walsh-Hadamard transform

A method is proposed for th efficient implementation of a class of second-o rder Volterra filters where their quadratic kernels are recursively constru cted from a set of isotropic subkernels (ISKs). The quadratic kernel of thi s class of filters can be factorised into a diagonal form by a Walsh-Hadama rd transform (WHT) and be implemented using a transform-and-multiply struct ure. An algorithm for determining the quadratic ISK kernel for the nonlinea r prediction of speech signals is presented.