摘要
为进一步提高神经网络模型的方向估计精度,提出利用特征矢量相角作为方向特征来构建模型。该方法首先通过协方差矩阵特征分解得到不易受噪声干扰的信号特征矢量;再对该矢量提取相角,信号的方向信息就包含在该相角中,以该相角作为输入矢量来训练模型。仿真结果证明了该方法具有抗噪能力强、模型估计精度高等特点,因此具有较高的工程应用价值。
In order to get the higher precision in neural network-based DOA estimation,this paper proposed a novel approach using phase-angle of eigenvector as the direction character.This approach firstly decomposed the covariance matrix to get the phase-angle of signal eigenvector,which contained DOA information with less noise interference,so the phase-angles of eigenvector could be used as the input vectors to train RBFNN model.The simulation results show this method has more powerful anti-noise ability and higher estimation precision,etc.so it has a good engineering application value.
出处
《计算机应用研究》
CSCD
北大核心
2011年第7期2655-2657,共3页
Application Research of Computers
基金
国家自然科学基金资助项目(60972161)
关键词
特征矢量相角
方向特征
来波方位估计
径向基神经网络
phase-angle of eigenvector
direction character
DOA estimation
RBFNN（radial basis function nearal networks）
