摘要
针对前馈神经网络盲均衡权系数随机初始化对接收信号能量的影响,利用能量匹配准则对前馈神经网络盲均衡算法的代价函数进行了修正,使算法在收敛过程中保证输入输出信号能量趋于一致,满足盲均衡实现的充要条件的前提成立,同时在修正代价函数下的前馈神经网络盲均衡算法具有更快的收敛速度,加强了算法的稳健性。计算机仿真证明修正代价函数的利用有效提高了前馈神经网络盲均衡的均衡性能。
The cost function is modified by energy-matching rule for blind equalization by FNN. This method overcomes the shortcoming of energy unsteadiness for weight of neural network random initialization, thus ensures the precondition of the necessary and sufficient condition for blind equalization. By using the modified cost function, blind equalization by FNNrepresents higher convergence rate and the algorithm is enhanced in robustness. The simulation results indicate that the modified cost function provides better performance for blind equalization by FNN.
出处
《通信技术》
2008年第12期115-116,119,共3页
Communications Technology
