迭代预测正交匹配追踪算法

发布时间：2019-04-03 08:24

【摘要】：正交匹配追踪系列算法中,每次迭代在原子库中选择和残差匹配的多个原子是主流的改进方向,但对多原子的选择标准却鲜有深入研究,一般是选择原子库中与残差相关系数中最大的K个原子,或者选择所有大于某一阈值的原子。本文以正交匹配追踪算法为原型,运用统计学方法,研究了相邻两次迭代中与残差相关系数最大的原子之间的关系,得出了其相关系数具有区间性的结论,这对一次迭代选择多个原子具有指导意义。该结论可以支撑对下一步迭代中的原子进行高概率预测。基于此,本文提出了迭代预测正交匹配追踪算法,实验结果表明,相对于其他匹配追踪算法,其在保证重构精度未降低的情况下,耗时有较大幅度降低。
[Abstract]:In a series of orthogonal matching tracking algorithms, the selection of multiple atoms in the atomic library and residual matching at each iteration is the main direction of improvement, but the selection criteria for polyatoms are rarely studied in depth. Generally, K atoms in the atom library are selected, or all atoms greater than a certain threshold value are selected. In this paper, based on the orthogonal matching tracking algorithm, the relationship between the atoms with the largest residual correlation coefficient in two adjacent iterations is studied by using statistical method, and the conclusion that the correlation coefficient is interval is obtained. This is instructive to select more than one atom in one iteration. This conclusion can support the high probability prediction of atoms in the next iteration. Based on this, an iterative predictive orthogonal matching tracking algorithm is proposed in this paper. The experimental results show that compared with other matching tracking algorithms, the time consuming of the iterative predictive orthogonal matching tracking algorithm is much lower than that of other matching tracking algorithms without decreasing the reconstruction accuracy.
【作者单位】：西安电子科技大学ISN国家重点实验室;河南农业大学机电工程学院;
【基金】：国家自然科学基金项目(61372069) 高等学校学科创新引智计划(111计划)资助项目(B08038) 河南省高等学校重点科研项目计划项目(15A510002)
【分类号】：TN911.7

【相似文献】