基于最小径向基函数(MRBF)神经网络的非线性噪声对消

Non-linear noise canceller based on minimal radial basis function neural networks

由于传统线性噪声对消器对非线性噪声不能很好的适应,本文改进传统线形噪声对消器的线形滤波器部分,用MRBF网络代替横向LMS结构,以适应非线形噪声存在的情况。使用改进的RBF网络结构MRBF[1],解决了隐单元数目确定的困难,学习算法采用更加稳定的扩展卡尔曼滤波方法(KEF),增强了RBF的实用性。本文提出的方法降低了运算复杂度,增强实时性。仿真结果表明该方法有良好的噪声对消性能。

As traditional linear noise canceller can not efficiently adapt to non-linear noise environment, we improved non-linear noise canceller by replace transverse LMS with MRBF network to enable the filter adapt non-linear noise environment. The improved MRBF network can efficiently conquer the difficulty of determine the hint unit number. A more steady algorithm extend Kalman filter （KEF） strengthen the applicability of RBF algorithm. Simulation test illustrate that the solution can dramatically reduce the operation complexity and achieve real time operation, and it has great performance in noise canceling.