一种新的阶递归平稳随机信号功率谱估计法

A New Estimation Method of Power Spectrum for Order Recurrence Stationary Random Signals

本文提出了一种新的平稳随机过程的功率谱估计方法。功率谱由自回归滑动平均(ARMA)模型参数来表示,在修正的Yule-Walker方程基础上产生嵌套的YW方程,采用格型结构确定AR参数并自动由AIC准则定阶。对分子施加最小相位限制后,再次利用格型公式求得MA分量的逆系统和阶数,最后用LPC技术确定分子参数。新方法克服了Cadzow法对相位灵敏的缺点,且使计算量大大下降,模拟结果表明,它具有更高的分辨率。

Our paper presents a new method of power spectrum estima- tion for stationary random signals. PSD is expressed by the parameters of an ARMA model. On the basis of the modified YW equation, we have deduce an embeded YW equation. AR coefficients are calculated out by the famous lattice structure and the order is determined automatically by AIC criteria, After transfomation, we use the same lattice algorithm to find MA parameters. The new method has overcome some disadvantages of Cadzow’s way such as negative spectrum, sensitivity to phase and large computation, etc. Several examples have de monstrated the higher resolutions of our algorithm.