摘要
提出了一种基于主分量分析(PCA)神经网络实现子快速子空间分解及其维数估计的新方法。该方法不需要估计数据协方差矩阵和特征值分解,只需将阵列数据输入到PCA网络,通过网络权值的无监督自组织迭代即可同时完成子空间分解及其维数估计。因此该方法具有运算量小和复杂度低的特点,易于实时处理。计算机仿真验证了该方法的有效性。
A fast method for subspace decomposition and its dimension estimation based on principal components analysis （PCA） network is proposed in this paper. This method can quickly estimate signal subspace via PCA network weight vectors by unsupervised self-organized learning rule, and need not involve any estimation of the covariance matrix or its eigendecomposition, thus indicating that the proposed method is computationally attractive and easy to implement in real time. Computer simulation results have also proved its effectiveness.
出处
《通信技术》
2009年第1期51-53,共3页
Communications Technology
关键词
子空间分解
主分量分析
人工神经网络
波达方向
subspace decomposition： principle components analysis
artificial neural network
direction-ofarrival
