摘要
利用支持向量回归机对非线性函数的拟合能力,将波达方向(DOA)估计问题转化为样本的智能学习问题。提取已知信号的协方差矩阵上三角部分作为样本输入特征,构建波达方向估计模型,获取复杂函数的拟合能力,达到对未知信号波达方向估计的目的。仿真实验表明该方法具有很高的估计精度和速度,在低信噪比和通道存在相位误差的情况下具有较强的适应能力,性能优于RBF神经网络法,具有较大的工程应用价值。
This paper transfers the problem of DOA estimation into a samples intelligent learning problem by using the approxima- ting capability of support vector regression for nonlinear functions . The upper triangular half of the covarianee matrix of know- ing direction signals is extracted to form training set which is used to construct DOA estimation model. The DOA estimation model can get the approximating capability for nonlinear functions to estimate the DOA. The experiment results show that the proposed method has a high estimation precision and speed,and has an advantage of preferably robust in the condition of low signal-to-noise and phase error in the channels. The performance is better than the RBFNN method and has a broad application future.
出处
《无线电工程》
2013年第2期34-37,共4页
Radio Engineering
关键词
协方差矩阵
支持向量机
来波方位
covariance matrix
support vector machine
direction of arrival
