关键词: 神经网络, 盲均衡, 常模算法

Abstract: In digital communications，the received signal is often corrupted by the inter-symbol interferences（ISI），and the ISI can be eliminated by blind equalization.The constant modulus algorithm is a widely applied blind equalization algorithm.For its slow convergence and big remnant errors，a new blind equalizer based on neural network is proposed.It can make the network convergent by very few training serial signals，and then the equalizer changes to the blind algorithm.The simulations show that this equalizer has less remnant errors and faster convergence speed than the ordinal equalizers，no matter in linear channel or nonlinear channels.

Key words: neural network, blind equalization, constant modulus algorithm