On the selection of memoryless adaptive laws for blind equalization in binary communications

摘要: We consider the adaptive equalization of an unknown linear time-invariant channel without observations input sequence, by updating impulse response coefficients equalizer with output times a memoryless nonlinear function output. To date, no such is known to result in global convergence inverse when consists binary data. The effect selection nonlinearity properties scheme studied, and it shown that for large class laws (including continuous functions), unequalized responses few nonzero are points convergence, there exist undesired local minima subset functions includes those previously proposed.