There are several methods for blind equalization of channels subject to linear inter-symbol interference and additive noise, among them those based on higher-order statistics or on Bussgang filtering. Under conditions of nonlinear ISI, these techniques will exhibit performance degradation which can range from mild to catastrophic in nature. We investigate this degradation as measured by convergence time and mean-squared error. We also look at an alternative architecture using a Bussgang filter in conjunction with a radialbasis function network for dealing with the nonlinear ISI. Finally, we compare the relative merits of each scheme, including ocnvergence rate, MSE at convergence, computational complexty and intialization procedures.
展开▼