单个正弦波下二维线性预测谱估计方法的改进

A New and Better Method for Estimating the Power Spectrum of Single Sine Wave Using the 2-D Linear Prediction(LP)

针对二维复正弦信号，提出了一种简单的线性预测的谱估计改进算法。其思想是利用可分离二维信号的性质，把单个二维复正弦信号分解成2个一维复正弦信号之积，然后分别对2个一雄信号进行谱估计。其间对一维线性预测误差滤波器系数进行适当修正，这样可得到较高分辨率和稳定性的正弦信号的谱估计结果。大量仿真结果表明，简单的修正可使二维信号功率谱第1峰更加尖锐，第2峰值比未修正前要小10-15dB。

Aim. We propose a new and better method. In sections 1 and 2 of the full paper, we explain our method in some detail. In this abstract, we just describe briefly it＇s two essential points： （1） using the property of the separable two-dimensional signal, we decompose the two-dimensional complex sine wave signal into the product of two one-dimensional complex sine wave signals; then we estimate respectively the two power spectra of the two one-dimensional complex sine wave signals ; （2） modifying the coefficient of the one-dimensional LP forecast error filter appropriately, we can gain higher resolution and better statistical stability of the power spectrum. Finally we did computer simulation, whose results, shown in Figs. 1 and 2 in the full paper, verify preliminarily that our new and better method can shrink the first peak of the two-dimensional signal power spectrum and diminish the second peak by 10-15 dB as compared with ordinary LP method.