CRIMNO: criterion with memory nonlinearity for blind equalization

A novel criterion with memory nonlinearity (CRIMNO) is introduced for blind equalization problems. The basic idea of CRIMNO is to make use of the fact that the transmitted data are statistically independent of each other. It is shown that CRIMNO may not have local minima if its weights are chosen properly, thereby guaranteeing global convergence. An adaptive weight CRIMNO algorithm is also presented and tested with simulation examples of quadrature amplitude modulation (QAM) signals. It is shown that the adaptive weight CRIMNO algorithm exhibits faster convergence speed than the Godard (1980) algorithm without any significant increase in computational complexity