ABSTRACT
In this chapter, the local convergence behaviors of various blind equalization algo rithms are addressed. The local convergences of Godard algorithm (GA), ShalviWeinstein algorithm (SWA), and a standard-cumulant algorithm (SCA) are first demonstrated for linear FIR equalizers. A geometric relationship between the Go dard algorithm (or the equivalent constant modulus algorithm) and the SWA is revealed. In fact, both algorithms share the same local or global convergence char acteristics. The specific convergence points of the SWA equalizer are shown to be influenced by a kurtosis condition which can be exploited for initializing equalizer parameters.