基于RBF神经网络与RLS算法的均衡器

Equalizer Based on RBF Neural Network and RLS Algorithm

将径向基函数神经网络与横向均衡器相结合,采用递推最小二乘算法更新权值。将最小二乘误差作为代价函数以及与误差相关的变步长,使输出误差较传统的神经网络均衡器进一步减小,收敛速度得到提高。仿真结果表明,该均衡器对线性信道和非线性信道都表现出较好的性能,在较严重的非线性情况下其优越性更明显。

This paper combines Radial Base Function（RBF） neural network and landscape filter, uses Recursive Least Square（RLS） algorithm to update the weight and uses variable steps associated with errors, the output error and the convergence speed are both improved. Simulations results show that the new equalizer has better performance, whether it is in linear or nonlinear. In more serious cases, its advantages are much more obvious.