Multiple-snapshot maximum-likelihood (ML) direction of arrival (DOA) estimation problem is studied for the intermittent jamming scenario. The intermittent jamming modality is based on the assumption that only a subset of the collected snapshots are contaminated by the jammer while the others are jammer-free; but the receiver does not know which is which. This type of jamming is frequently encountered in practice either inadvertently, say due to the sporadic activity of a non-hostile system; or intentionally, say due to the activity of an adversary sweeping the operational bandwidth of the receiver. Exact maximum likelihood solution for the problem is analytically intractable and an expectation maximization (EM) method based solution is developed for coherent and non-coherent signal models. Coherent signal model assumes that the phase difference between the coefficients of two consecutive snapshots are known a-priori which is an assumption compatible with the Swerling-1/3 target models in the radar signal processing literature. Non-coherent signal model does not have such an assumption and it is suitable for Swerling-2/4 targets. The suggested EM based solution is shown to yield an important estimation accuracy improvement over conventional maximum-likelihood solution which ignores the intermittency of jammer and also over the atomic norm based high resolution estimation techniques. Cramer-Rao type performance lower bounds for the problem is also provided to illustrate the efficacy of the suggested estimator.

针对间歇性干扰场景，研究了多快照最大似然（ML）到达方向（DOA）估计问题。间歇性干扰方式基于以下假设：收集的快照中只有一部分受到干扰者的污染，而其他快照没有干扰。但是接收者不知道哪个是哪个。实际上，这种干扰通常是由于非敌对系统的零星活动而无意中遇到的。或有意地说（例如，由于敌方的活动扫荡了接收机的操作带宽）。该问题的确切最大似然解在分析上难以解决，并且针对相干和非相干信号模型开发了基于期望最大化（EM）方法的解决方案。相干信号模型假设两个连续快照的系数之间的相位差是先验的，这是与雷达信号处理文献中的Swerling-1 / 3目标模型兼容的假设。非相干信号模型没有这样的假设，它适用于Swerling-2 / 4目标。与传统的最大似然性解决方案相比，该解决方案显示出建议的基于EM的解决方案可以显着提高估计精度，传统的最大似然性解决方案忽略了干扰的间歇性以及基于原子范数的高分辨率估计技术。还提供了该问题的Cramer-Rao型性能下界，以说明建议的估计量的有效性。