A novel radar high-resolution range profile (HRRP) target recognition method based on a stacked autoencoder (SAE) and extreme learning machine (ELM) is presented in this paper. As a key component of deep structure, the SAE does not only learn features by making use of data, it also obtains feature expressions at different levels of data. However, with the deep structure, it is hard to achieve good generalization performance with a fast learning speed. ELM, as a new learning algorithm for single hidden layer feedforward neural networks (SLFNs), has attracted great interest from various fields for its fast learning speed and good generalization performance. However, ELM needs more hidden nodes than conventional tuning-based learning algorithms due to the random set of input weights and hidden biases. In addition, the existing ELM methods cannot utilize the class information of targets well. To solve this problem, a regularized ELM method based on the class information of the target is proposed. In this paper, SAE and the regularized ELM are combined to make full use of their advantages and make up for each of their shortcomings. The effectiveness of the proposed method is demonstrated by experiments with measured radar HRRP data. The experimental results show that the proposed method can achieve good performance in the two aspects of real-time and accuracy, especially when only a few training samples are available.

Radar high resolution range profiles(HRRP) can represent abundant target structure signatures, which has found wide-spread applications. Recently, radar automatic target recognition(RATR) methods based on deep neural network have achieved promising results because of their strong generalization ability. However, these deep models usually require a large amount of labeled data to optimize the parameters, otherwise they would possibly encounter severe overfitting problem in few-shot condition. In order to solve the above problem, a novel few-shot HRRP target recognition method based on meta-learning framework is proposed, which introduces Long Short-Term Memory(LSTM) based neural network as learner for statistical HRRP data. The proposed method exploits multi-polarization HRRP data for RATR and successfuly improves recognition accuracy and generalization performance in few-shot condition. In the proposed method, a novel learner is designed, which is more suitable for processing one-dimensional statistical HRRP data. We evaluated our method based on an electromagnetic calculation dataset of airplanes and found that the proposed method could successfully fuse multi-polarization HRRP data to provide more effective information for RATR. The experimental results also showed that the proposed method produced improved performance compared with state-of-the-art few-shot learning methods.

The ambiguity function is a useful tool for assessing the response performance of radar waveforms. This paper addresses the problem of shaping the radar ambiguity function by designing the unimodular transmit waveform. We model the waveform design process as a non-convex quartic optimization problem aimed at minimizing the disturbance power of scatters in specific range-Doppler bins. Rather than using conventional methods like approximating and relaxing the objective function and constraints, we address the resultant optimization problem under the Riemannian manifold optimization framework and introduce a novel gradients-based algorithm called Quartic Riemannian Adaptive Regularization with Cubics. The developed algorithm applies the Riemannian gradient and Riemannian Hessian of the objective function to conduct iterative operations that gradually decrease the cost function value. The proposed algorithm has a lower iteration complexity bound in comparison to the classical Riemannian trust region method based on the second-order gradient. Numerical experiments demonstrate that our developed method can achieve a relatively higher signal-to-interference ratio with a faster convergence speed.

In practice, adaptive beamforming usually faces non-ideal situations where a limited number of snapshots are available, the training data are corrupted by desired target signals and the array mismatches exist. Traditional methods often degrade significantly under the above situation. In order to solve this problem, a new adaptive beamforming method based on atomic-norm optimization technique is proposed in this paper. In the proposed method, the interference subspace is estimated by minimizing the rank of interference data matrix while making the signals bounded within a ball of Frobenius norm around the observed data. This non-convex problem is solved using alternative optimization which decomposes it into two iterative steps. Each step can be formulated as semi-definite programming and solved efficiently. Unlike traditional methods, the proposed method can estimate the target signals, target directions and interference subspace simultaneously. This property guarantees that the proposed beamformer is free from the influence of target signals and able to adjust pointing direction adaptively. Then it is derived theoretically that the estimation of interference subspace in the proposed method is consistent and bounded. A fast implementation algorithm based on alternating direction method of multipliers is also derived. Compared with traditional methods, the proposed method not only performs better with target-contaminated training data and erroneous prior of target direction, but also requires much less snapshots to work. The effectiveness of the proposed method and its advantages over traditional methods are verified based on simulated and actual measured radar data.

Waveform adaptation grants cognitive radar (CR) the ability to adapt to its environment, which requires an effective framework to synthesize waveforms sharing a desired ambiguity function (AF). In this letter, we propose a novel method for shaping the slow-time AF in order to adaptively suppress the interference power. The problem is formulated as minimizing the average value of the slow-time AF over some range Doppler bins spanned by the interference, which can be identified exploiting a plurality of knowledge sources. From a technical point of view, this is tantamount to optimizing a complex quartic-order polynomial with a constant modulus (CM) constraint on each optimization variable. To solve this problem, we proposed a quartic Riemannian trust region algorithm. This algorithm first transforms the optimization into an unconstrained one in a complex circle Riemannian manifold, then devises a new Riemannian trust region optimization algorithm that invokes Riemannian gradient and Hessian matrix to obtain an iterative solution with super-linear convergence rate and ability to escape potential saddle points. Simulation results demonstrated that our proposed algorithm outperformed state-of-the art approaches for AF shaping while being computationally less expensive.

Direction-of-arrival (DOA) estimation using sparsity-inducing techniques has attracted much interest recently. In this paper, the DOA estimation for the bi-static passive radar is investigated. Under the framework of sparse Bayesian learning (SBL), a joint sparse Bayesian model is established to combine the measurements from both stations and yield improved targets DOA estimation. First, the maximum a posteriori (MAP) estimation of the DOA using the joint data set is derived. With the utilization of more measurements, the joint reconstruction process can produce far more precise estimates. To reduce the computational expense, a fast SBL method based on evidence maximization is also proposed. Using the fix-point method, the fast SBL method tends to converge faster than the MAP estimator. Theoretical results focusing on local convergence property of the fast SBL method are provided. The simulation results show that the proposed methods outperform the conventional SBL methods, especially in harsh scenarios with a limited number of snapshots and low signal-to-noise ratio (SNR).