论文部分内容阅读
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。
Aiming at the effect of random equalization of blind equalization weight of pre-fed neural network on the received signal energy, the cost function of the blind equalization algorithm of feedforward neural network is modified by the energy matching criterion so that the energy of the input / output signal tends to be convergent The premise of satisfying the necessary and sufficient condition of the realization of the blind equalization is established. Meanwhile, the blind equalization algorithm of the feedforward neural network with the modified cost function has a faster convergence speed and strengthens the robustness of the algorithm. The computer simulation proves that the use of the modified cost function effectively improves the equalization performance of the blind equalization of feedforward neural networks.