基于稀疏迭代协方差估计的缺失数据谱分析及时域重建方法

发布时间：2019-06-22 11:06

【摘要】：应用于缺失数据恢复的迭代自适应方法(IAA)被证实可利用20%的有效数据估计信号参数,并能高精度恢复缺失数据,优于经典GAPES方法,但当缺失数据超过80%时其数据恢复性能迅速下降。该文基于稀疏迭代协方差估计提出一种新的缺失数据谱分析方法(M-SPICE)及针对该方法的缺失数据修正时域重建方法。该方法将加权缺失数据协方差拟合代价函数转换为凸优化问题,构造循环最小化器保证缺失数据参数估计的全局收敛特性,通过对缺失数据估计算子的更新实现了时域重建方法的修正,使其在有效数据功率谱欠估计的情况下获得更高的数据重建精度。仿真实验表明无论是数据块缺失还是任意缺失,该方法均能够利用更少的有效数据进行谱分析,并重建大比例缺失数据。
[Abstract]:The iterative adaptive method (IAA) for missing data recovery has been proved to be able to estimate signal parameters by using 20% effective data, and can recover missing data with high precision, which is superior to the classical GAPES method, but its data recovery performance decreases rapidly when the missing data exceeds 80%. Based on sparse iterative covariance estimation, a new missing data spectrum analysis method (M-SPICE) and a modified time domain reconstruction method for missing data are proposed in this paper. In this method, the weighted missing data covariance fitting cost function is transformed into a convex optimization problem, and a cyclic minimizer is constructed to ensure the global convergence of missing data parameter estimation. The time domain reconstruction method is modified by updating the missing data estimation operator, so that it can obtain higher data reconstruction accuracy when the effective data power spectrum is underestimated. The simulation results show that the method can use less effective data for spectral analysis and reconstruct a large proportion of missing data, whether it is missing data or arbitrary missing data.
【作者单位】：北京理工大学电子与信息学院;中国人民解放军军械工程学院电子与光学工程系;
【基金】：国家自然科学基金(61401024)~~
【分类号】：TN911.7

【相似文献】