Attention-Based Radar PRI Modulation Recognition With Recurrent Neural Networks

Abstract

Analyzing radar signals is a critical task in modern Electronic Warfare (EW) environments. However, the pulse streams emitted by radars have flexible features and complex patterns which are difficult to be identified from a statistical perspective. To solve this problem, pulse repetition interval (PRI) is used as a distinguishing parameter of emitters to be identified. However, traditional PRI modulation recognition methods can only deal with simple PRI modulations and their performance will further degrade with the increasing number of emitters or noisy environments. In this paper, we introduce an attention-based recognition framework based on recurrent neural network (RNN) to categorize pulse streams with complex PRI modulations and in environments with high ratios of missing and spurious pulses. Simulation results show that our model is robust to noisy environments and has a better performance than conventional methods. INDEX TERMS Attention mechanism, electronic warfare, PRI modulation, recurrent neural network (RNN).

Multi-attribute overlapping radar working pattern recognition based on K-NN and SVM-BP

“…Paper on radar working recognition mainly divides radar working mode into four typical working: speed search, ranging while searching, scanning while tracking, and single target tracking [1]. For example, [26][27][28][29] (2,2) num r (2,3) num ⋯ r (2,N) num r (3,2) num r (3,3) num ⋯ r (3,N) num ⋮ ⋮ ⋱ ⋮ r (N,2) num r (N,3) num ⋯ r (N,N) num ⎤ ⎥ ⎥ ⎥ ⎦ side's disability to assess accurately the status of the enemy aircraft in air combat and even miss the aircraft.…”

“…Electronic war (EW) has developed into an independent mode of operation with high-tech in modern warfare. It is often the beginning of a war and can be decisive to the victory [1]. The recognition of radar working patterns is one of the main components of current cognitive EW.…”

Very Short-Term Power System Frequency Forecasting

“…We follow the approach described in [35] to incorporate the frequency measurement errors to the proposed methodology. To this end, we augment Dataset I training and validation sets with data obtained by adding an error term that follows the distribution N 0, (1.667)10 −3 to each measurement reported in Dataset I training and validation sets.…”

“…We train the model with the loss function (6) using the 2N samples of the augmented Dataset I training set.We similarly follow the approach in [35] to evaluate the performance of the models. We modify the two test sets by adding an error term that follows the distribution N 0, (1.667)10 −3 to each measurement in Dataset I test set and Dataset II and refer to the resulting test sets as modified Dataset I test set and modified Dataset II, respectively.…”

Synthetic algorithm for deinterleaving radar signals in a complex environment

“…The PRI parameter has attracted increasing attention. Identifying the PRI modulation patterns hidden in it has become a new research direction [20–22]. With the development of artificial intelligence, support vector machine (SVM) [23], fuzzy clustering [24], and neural network [25–27] technology are applied to deinterleaving.…”