基于自适应共轭梯度算法的高分辨率谱估计器

High resolution spectral estimator based on adaptive conjugate gradient algorithm

基于AR模型的功率谱估计是现代谱估计应用非常广泛的一种方法。通过对一全极点模型的参数估计来实现功率谱估计。提出了一种采用自适应共轭梯度算法来进行参数估计的方法。由于共轭梯度算法采用迭代运算求解Yule-Walker方程,同现有的谱估计算法相比,大大减小了谱估计算法的计算复杂度;随着自相关矩阵阶数的增大,该方法谱估计精度在小信噪比下提高显著。仿真结果表明,这种方法和基于AR模型的其它谱估计方法在不同信噪比下具有几乎相同的分辨率。因此,该谱估计算法具有重要的实用意义,有助于谱估计算法的实时实现。

Spectral estimation algorithms based on AR model are widely used in modem spectral estimation. In general, spectral estimation is implemented through the parameter estimation of an all pole model. There are many solutions to get these parameters. An adaptive conjugate gradient algorithm for the parameter estimation to realize spectral estimation is proposed. Since this algorithm uses recursive technique to solve Yule-Walker equation, it leads to the great reduction of computation complexity with comparison to the existing algorithms. As the dimension increase of the covariance matrix, the improvement of resolution under the low SNR is very obvious. The simulation results show that the proposed spectral estimation algorithm has almost the same resolution as compared with other traditional methods under all kinds of SNR conditions. Therefore it can be concluded that this algorithm is more practical and suitable to real time applications.