SOFT CONSTRAINT SATISFACTION (SCS) BLIND CHANNEL EQUALIZATION ALGORITHMS

The constant modulus adaptive blind equalization algorithms presented in this paper are shown to correspond to an error performance surface which is much improved upon that of existing algorithms such as the we ll-known constant modulus (or Godard) algorithm. Many undesirable loca l solutions (ULSs) are avoided by careful derivation. We use a determi nistic optimization criterion with a soft constraint to obtain an upda te equation which contains a normalized gradient vector and a particul ar continuous non-linearity. This approach is extended to multiple con straints to yield faster converging algorithms. An autoregressive (AR) channel model is studied to demonstrate analytically the absence of a class of ULSs. Finally, the findings are verified experimentally for various AR and moving-average (MA) channels. (C) 1998 John Wiley & Son s, Ltd.