摘要
提出了一种改进的基于递归神经网络的盲均衡算法，它利用信号的高阶统计特性构造了代价函数，用共轭梯度算法对递归神经网络进行训练，且利用线性搜索法对参数进行动态选取，模拟结果显示该算法能很好的用于各种信道及信号的均衡．
A new recurrent neural network based blind equalization of algorithm is proposed. Its cost function is based on higher order statistics of input signals, and the conjugate gradient algorithm is used in neural network training. Besides, the parameter values are chosen by linear searching algorithm. Simulization results show that our algorithm can fit for different kinds of channels and signals.
出处
《北京邮电大学学报》
EI
CAS
CSCD
北大核心
1997年第4期6-11,共6页
Journal of Beijing University of Posts and Telecommunications
基金
国家自然科学基金
关键词
神经网络
代价函数
盲均衡算法
信道均衡
neural networks
cost function
linear search
blind equalization algorithm
