【摘要】 现代互谱估计是抑制有色观测噪声的一种有效方法.本文针对旋转不变技术来估计信号参量的方法进行了深入的分析,进而提出了在有色观测噪声背景下,基于最小二乘技术的信号估计的互谱ESPRIT方法。这种方法的突出特点是不需要任何色噪声的先验信息准确地估计出待估计信号,不仅避免了以往互谱估计本身所固有的在整个频域上的谱峰搜索,而且克服了自谱ESPRIT估计方法对噪声的严格限制(要求噪声为白色高斯噪声)。可直接通过特征值确定信号参数估计值。本文方法只需两次SVD分解,计算量大大降低。仿真结果表明,该方法具有较好的谱估计分辨率和良好的稳定性。更多还原

【Abstract】 Modern cross spectrum estimation is an effective method to restrain coloured observation noise.By using further analysis of signal estimation for parameters via rotational invariance techniques,this paper proposes a cross-spectral ESPRIT method based on least squares technique in colored stochastic noise.The mainly advantage of this method is that signals can be accurately estimated without any priori information of colored noises,which overcomes the searching procedure inherent in all the previous cross-spectral estimation and the limitation of auto-spectral estimation methods for noise(requires noise to be white and gauss) and signal parameter estimation can be directly determined by characteristic value.Two times of SVD decomposition makes the amount of computation greatly deduced.Simulation result shows that this method exhibits good spectral estimating differentiability and stability.更多还原