基于径向基函数神经网络的波束形成算法

Beamforming algorithm based on RBF neural network

提出一种基于径向基函数(RBF)神经网络的波束形成算法。针对最小方差无失真响应(MVDR)算法,由接收信号的协方差矩阵计算其权矢量;将协方差矩阵以列向量的形式输入RBF神经网络,对其加以训练,使之逼近MVDR算法的权矢量;将训练好的RBF神经网络用于波束形成中,对不同角度的接收信号,RBF神经网络可自适应地输出相应权矢量。仿真结果表明,基于RBF神经网络的波束形成算法能快速逼近任意波束算法的权矢量,波束赋形效果良好,与已有波束形成算法相比,可降低算法复杂度,减少计算量。

A beamforming algorithm based on radial basis function （RBF） neural network is presented in this paper. According to the minimum variance distortionless response （MVDR） algorithm, the weight vector is calculated by the covariance matrix of the received signal; the covariance matrix is put into RBF neural network in the form of column vector, and Rt3F neural network is trained in order to approximate to the weight vector of MVDR algorithm. The trained RBF neural network is then used in the beamforming. The RBF neural network can be adaptive to output the corresponding weight vector according to the received signal from different angles. Simulation results show that the new algorithm can fast-approximate to weight vectors of any beamforming algorithm. The effect is good compared with that of the existing beamforming algorithm. The new algorithm can reduce the complexity of the algorithm and reduce the amount of calculation.