非线性信道判决反馈递归神经网络自适应均衡器

针对存在严重符号间串扰及轻度非线性畸变的数字信道，提出了一种判决反馈递归神经网络为均衡器，它将传统的线性信道判决反馈结构巧妙地融入递归神经中，在自适应训练时用期待２信号代替判决反馈信号。

A DECISION FEEDBACK RECURRENT NEURAL NETWORKS EQUALIZER FOR NON LINEAR CHANNELS

A decision feedback equalizer employing recurrent neural networks structure used for non linear channels with severe linear distortion and mild non linear distortion is proposed in this paper.We put the recurrent neural network skilfully into the traditional decision feedback structure for linear channels equalization and substitue trainning signal instead of decision feedback signal in the learnig process,and performance better than that when recurrent neural network equalizer alone is embed in the system is obtained when the number of parameters are equal in both structures.Simulative results have shown that the proposed equalizer is simple,convenient in real time signal processing and the equalization performance is improved significantly.