Randomized Low-Rank and Sparse Decomposition-Based Receiver Search Strategy for Electronic Support Receivers

Narrow bandwidth electronic support (ES) receivers have higher sensitivity than wide bandwidth ES receivers. However, to intercept the emitting signals in a wide spectrum range, the frequency spectrum should be swept over time. Hence, a scanning strategy has to be used. Mostly, the strategy is determined based on prior information about possible threats. Hence, there is a lack of robustness to determine the best scanning strategy. With using only the prior information, the policies cannot model a closed-loop system. These types of strategies are insufficient to use the advantage of previously gathered data. At this point, we propose a method to determine the scanning strategy of the frequency spectrum, which learns each emitting signal pattern by using transformed predictive state representations. This method benefits from the recent advances in a rank minimization problem using the GoDec algorithm and allows us to use low-rank and also sparse component to predict the best possible frequency band. With this method, time consumption of the most hotspot part of the recursive process, the subspace identification part is decreased around 80$\%$, and also better average radar detection ratio are achieved.