RBF神经网络多用户检测ROLS-AWS算法的研究

为了对抗多址干扰和远近效应，研究将RBF(经向基函数，Radial Basis Function)神经网络中的递归正交最小二乘(ROLS-AWS)算法应用于多用户检测中。给出在同步高斯信道条件下运用三层神经网络解调扩频信号的原理框图，分析了基于RBF网络的多用户检测接收机。为了改进RBF网络的运算速度，在基于RBF网络的多用户检测接收机中采用ROLS-AWS算法。计算机仿真结果表明：使用所提算法的RBF网络接收机的抗多址干扰、远近效应以及训练速度的性能上都明显优于传统接收机、基于BP神经网络和不使用该算法的普通RBF神经网络多用户接收机。

ROLS-AWS Algorithm Used in RBF Neural Network Multiuser Detection

In order to suppress the multiple access interference (MAI) and resist near-far effect, the recursive orthogonal least square with auto weight selection(ROLS-AWS) algorithm used in radial basis function(RBF) neural network is introduced to the multiuser detection(MUD).The paper first introduces RBF into MUD. Then the three-layer neural network demodulation spread-spectrum signal model in synchronous Gauss channel was given. The multi-user detection receiver was analyzed. In order to improve the computational speed, the ROLS-AWS algorithm was used in the RBF-based MUD receiver. The simulated results show that the proposed RBF-based MUD receiver using ROLS-AWS algorithm is better than the conventional detector, the common BP and the RBF neural network which does not use ROLS-AWS based MUD receiver on suppressing multiple access interference,near-far effect and training speed.