摘要
分析了现有的基于最小二乘法的AR参数模型的谱估计算法在信噪比较低时估计效果差的原因,提出了一种基于协方差成形最小二乘法的改进的参数模型AR谱估计算法。这种算法建立了以线性模型的真实输出与估计输出的均方误差为模型的代价函数,并选择满足一定约束条件的线性变换估计使得该均方误差最小。仿真结果表明,这种算法虽然是有偏估计,但在信噪比不高的情况下,估计效果优于Yule Walker等参数模型AR谱估计方法,而在信噪比较高的情况下,二者估计效果相当。
This paper proposes a modified parametric method of AR power spectrum estimation based on covariance shaping least squares, and shows the reason for the poor estimate of the existing parametric method which is based on leastsquares when the SNR is low or moderate. At first we build a cost function which is the MSE between the real output and the estimate output of a line model, next we select a line transform is optimal in the sense that results in the estimate output is as close as possible to the real output in MSE. The emulate result shows that the new method can significantly outperforms the existing parametric method such as YuleWalker when the SNR is low to moderate, and if the SNR is high, the new method can get the approximately result as the latter.
出处
《现代电子技术》
2005年第8期85-86,96,共3页
Modern Electronics Technique
关键词
自回归
协方差成形
最小二乘
谱估计
auto-regress
covariance shaping
least-squares
power spectrum estimation