Target-Aware Recurrent Attentional Network for Radar HRRP Target Recognition

doi:10.1016/j.sigpro.2018.09.041

Signal Processing

Volume 155, February 2019, Pages 268-280

https://doi.org/10.1016/j.sigpro.2018.09.041 Get rights and content

Highlights

•
RNN can learn discriminative HRRP features by considering the temporal correlation.
•
Spectrogram containing phase information is more suitable for HRRP ATR.
•
The attention mechanism can automatically focus on the discriminative target areas.
•
Combining attention mechanism and RNN alleviates the time-shift sensitivity of HRRP.

Abstract

In this paper, we develop a Target-Aware Recurrent Attentional Network (TARAN) for Radar Automatic Target Recognition (RATR) based on High-Resolution Range Profile (HRRP) to make use of the temporal dependence and find the informative areas in HRRP, since it reflects the distribution of scatterers in target along the range dimension. Specifically, we utilize RNN to explore the sequential relationship between the range cells within a HRRP sample and employ the attention mechanism to weight up each timestep in the hidden state so as to discover the target area, which is more discriminative and informative. Effectiveness and efficiency are evaluated on the measured data. Compared with traditional methods, besides the competitive recognition performance, TARAN is also more robust to time-shift sensitivity thanks to the memory function of RNN and attention mechanism. Furthermore, detailed analysis of TARAN model are provided based on time domain and spectrogram features.

Introduction

A High-Resolution Range Profile (HRRP) is composed of the amplitude of the coherent summations of the complex returns from target scatterers in each range cell, which represents the projection of the complex returned echoes from the target scattering centers onto the radar Line-Of-Sight (LOS), as shown in Fig. 1. Since it contains abundant discriminative information, such as target size, scatterer distribution, radar HRRP target recognition has received intensive attention from the Radar Automatic Target Recognition (RATR) community [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19].

For the RATR task, feature extraction is a critical step [20]. Liao et al. [21] extract invariant features based on the integration of the bispectra of range profiles. Du et al. [22] further investigate the recognition methods based on high-order spectra features. In [23], the feasibility of classifying aerial targets using the micro-Doppler signatures is studied, where the features are computed in the form of bicoherence estimates and cepstral coefficients. Although those algorithms based on engineered features are effective for recognition, the features are hand-crafted and rely on the personal experiences.

In recent years, deep learning algorithms have been explored for RATR. The autoencoder models have been improved to extract robust features for HRRP [13], which employ the average profile of each HRRP frame as the correction term to establish the cost function under the Mahalanobis distance criterion, and achieve better recognition performance compared with traditional Stacked Auto-Encoders (SAE) [24]. To model the aspect sectors with few HRRP data, Pan et al. [25] utilize the Deep Belief Network (DBN) to learn discriminative features by sharing latent information of HRRP data globally. However, those networks do not explore the temporal dependence between range cells, although they record the structure information of targets.

Several approaches have been proposed to describe the temporal dependence in HRRP. Pan et al. [15] transform the HRRP sample into a sequence and model the structure across range cells by the Hidden Markov Models (HMM) with a transition probability. To model the HRRP sequence, [4] describes the spatial structure across range by the hidden Markov structure and models the temporal dependence between HRRP samples by the time evolution of the transition probabilities. Instead of utilizing the HMM for RATR, it is possible to employ the Recurrent Neural Network (RNN) which exploits larger state-space and richer dynamics compared to HMM [26]. RNN is a class of artificial neural network with connections between hidden units, which allows the model to exhibit dynamic temporal behavior in the data [27]. It achieves state of the art performance on sequential data for different tasks, such as language modeling [28], speech recognition [26], machine translation [29] and so on.

In this paper, we take advantage of the temporal dependence within each HRRP sequence to use RNN with voting strategy to learn its latent representation for recognition task, where each HRRP is divided into multiple overlapping sequential feature as input. Furthermore, we integrate the basic formulation of RNN with an attention mechanism to reduce the influence of the noise areas and automatically catch the discriminative target areas in HRRP, called Target-Aware Recurrent Attentional Network (TARAN). The label for a HRRP sample is decided based on a weighted sum of the output from each hidden state computed by a multilayer perceptron. Therefore, different from the conventional feature selection methods [31], [32], TARAN aims at evaluating the importance of input at each timestep for recognition rather than individual features and the learned corresponding weights are data-dependent rather than fixed for all of data, which may have different lengths for different HRRP samples.

Inspired by [33], besides the HRRP features in time domain, we also try spectrogram features as input, which is a two-dimensional feature providing the variation of frequency domain with time domain.

The remainder of the paper is organized as follows. We formulate the RATR task with RNNs in Section 2. The paper introduces the proposed architecture, i.e., TARAN for radar HRRP target recognition in Section 3. The parameter setting and detailed experimental results based on measured data are made in Section 4, followed by conclusions in Section 5.

Section snippets

Problem formulation

We formulate the problem in a generic manner, and consider classification of an object into one of K distinct classes based on a D-dimensional data measurement of that object. Specifically, given a set of N input signals $X = [x (1), \dots, x (N)] \in R^{D \times N},$ where X is the training data set containing N input signals, D is the dimensionality of each signal.

Denote $y = [y (1), \dots, y (N)] \in R^{K \times N},$ where y(n) denotes the 1-of-K representation of the label of x(n), i.e., y(n) ∈ {0, 1}^K, $y_{k} (n) = 1$ if the label of x(n) is k, and

Proposed model

To apply the RNN model, the input signals need to be first converted into sequential features. In this section, we consider two kinds of sequential features i.e. time sequential feature and spectrogram feature.

Experimental results

To show the effectiveness of the proposed TARAN model, we will compare our models to other counterparts on the measured data according to different criterions in the experiments. We will first briefly introduce our measured radar HRRP dataset and the detailed experiment settings. In the experiments, we will discuss the influence of different parameters on our model and give the detailed analysis based on the quantitative recognition performance. To provide qualitative analysis, we will look

Conclusions

In this article, RNNs are analyzed and utilized for HRRP-based RATR to further take advantage of the temporal dependence between range cells. In order to find the discriminative areas in HRRP, we further propose Target-Aware Recurrent Attentional Network (TARAN), which employs the RNN to encode the HRRP sequence into hidden states and weight the hidden state at each timestep to focus on the discriminative regions data-dependently with the help of attention strategy. All the pieces of the TARAN

Acknowledgment

This work is partially supported by the Thousand Young Talent Program of China, NSFC (61771361), 111 Project (B18039), and the National Science Fund for Distinguished Young Scholars of China (61525105).

References (47)

L. Du et al.
Robust statistical recognition and reconstruction scheme based on hierarchical bayesian learning of HRR radar target signal
Expert Syst. Appl.
(2015)
H. Liu et al.
Radar high-resolution range profiles target recognition based on stable dictionary learning
Iet Radar Sonar Navig.
(2016)
B. Feng et al.
Radar HRRP target recognition with deep networks
Pattern Recognit.
(2017)
J.L. Elman
Finding structure in time
Cogn. Sci.
(1990)
B. Kingsbury et al.
Robust speech recognition using the modulation spectrogram
Speech Commun.
(1998)
S. Amari
Backpropagation and stochastic gradient descent method
Neurocomputing
(1993)
B. Chen et al.
Large margin feature weighting method via linear programming
IEEE Trans. Knowl. Data Eng.
(2009)
X.D. Zhang et al.
A new feature vector using selected bispectra for signal classification with application in radar target recognition
Signal Process. IEEE Trans.
(2001)
L. Du et al.
Bayesian spatiotemporal multitask learning for radar HRRP target recognition
IEEE Trans. Signal Process.
(2011)
J. Chai et al.
Combinatorial discriminant analysis: supervised feature extraction that integrates global and local criteria
(2009)

X. Liao et al.

Identification of ground targets from sequential high-range-resolution radar signatures

IEEE Trans. Aerosp. Electron. Syst.

(2002)

L. Du et al.

A noise-robust radar target classification method based on complex probabilistic principal component analysis

General Assembly and Scientific Symposium

(2014)

L. Shi et al.

Radar HRRP statistical recognition with local factor analysis by automatic bayesian ying-yang harmony learning

IEEE Trans. Signal Process.

(2011)

L. Du et al.

Radar HRRP target recognition based on dynamic multi-task hidden markov model

Proceeding of Empirical Methods in Natural Language Processing (EMNLP)

(2011)

B. Feng et al.

Radar HRRP target recognition based on k-SVD algorithm

IEEE Cie International Conference on Radar

(2012)

B. Feng, L. Du, C. Shao, P. Wang, H. Liu, Radar HRRP target recognition based on robust dictionary learning with small...

L. Du et al.

A two-distribution compounded statistical model for radar HRRP target recognition

IEEE Trans. Signal Process.

(2006)

M. Pan et al.

Multi-task hidden markov model for radar automatic target recognition

IEEE Cie International Conference on Radar

(2011)

L. Du et al.

Radar automatic target recognition based on complex high-resolution range profiles

International Conference on Radar

(2006)

C. Wang et al.

Automatic target recognition in high resolution sar image based on backscattering model

Dianbo Kexue Xuebao/Chin. J. Radio Sci.

(2004)

M. Pan et al.

Multi-task hidden markov modeling of spectrogram feature from radar high-resolution range profiles

EURASIP J. Adv. Signal Process.

(2012)

R.A. Mitchell et al.

Robust statistical feature based aircraft identification

IEEE Trans. Aerosp. Electron. Syst.

(1999)

D. Bahdanau et al.

Neural machine translation by jointly learning to align and translate

Int. Conf. Learn. Represent.

(2015)

Cited by (79)

SCNet: Scattering center neural network for radar target recognition with incomplete target-aspects
2024, Signal Processing
Most of the previous radar automatic target recognition (RATR) methods based on high resolution range profile (HRRP) are designed under the assumption of complete target-aspects, which assumes the HRRP samples with different target-aspects are complete in training dataset or template library. Few works were concentrated on HRRP RATR with incomplete target-aspects. However, it is extremely difficult and sometimes even impossible to obtain HRRPs with complete target-aspects in real world applications. Therefore, it is required to recognize targets of unseen target-aspects with an incomplete target-aspect template library. Aiming at this problem, a scattering center neural network (SCNet) is proposed for radar HRRP target recognition with incomplete target-aspects. Based on the assumption that an HRRP sample can be represented by a linear combination of a few atoms from a scattering center dictionary, we proposed a scattering center layer (SC-layer), which encapsulates the dictionary coding into an implicit layer and treats the scattering center dictionary as its parameter that is optimized in an end-to-end manner. We further proposed a discriminative target-aspect frame dictionary by dividing HRRPs of each class into multiple target-aspect frames and associating frame label information with each atom of it to enforce discriminability in features. A multiple target-aspect prototype classifier is proposed, which is more robust to intra-class variations and thus more suitable to handle the incomplete target-aspects problem compared with softmax classifier. The classification is simply implemented by matching HRRP features with each prototype and each HRRP is assigned to the class with the nearest prototype. In order to integrate both intra-class compactness and inter-class separation, we proposed a novel discriminative prototype loss. Taking the advantage of the discriminative prototype loss, HRRPs belonging to the same class and the same target-aspect frame are pulled closer to the corresponding prototype in the feature space, while simultaneously pushing apart from other prototypes of different classes. Experiments on the aircraft electromagnetic simulation dataset and the measured dataset demonstrated the superior performance of the proposed method compared with other HRRP-based RTAR methods under the condition of incomplete target-aspects.
CVPCNN: Conditionally variational parameterized convolution neural network for HRRP target recognition with imperfect side information
2024, Signal Processing
Radar target recognition tasks often incorporate valuable side information (SI) that aids in the recognition task. The effectiveness of SI has been fully verified in traditional statistical modeling approaches. Its full potential remains untapped in deep learning-based high-resolution range profile (HRRP) target recognition tasks. Taking into account the azimuth sensitivity of HRRP, azimuth is chosen as the SI of the proposed model. To integrate SI into deep neural networks, we propose two novel parameterized convolution methods, namely variational parameterized convolution (VPCONV) and the extended conditionally VPCONV (CVPCONV). Specifically, VPCONV makes the convolutional kernel weights as the function of the auxiliary azimuth, which gives the network the capability of adaptively adjusting to changes in azimuth. Acknowledging the challenge of estimating HRRP azimuth accurately in practical scenarios, VPCONV employs a heteroscedastic Gaussian distribution, featuring varying mean and variance, to generalize the estimation error of SI through variational encoding. Moreover, CVPCONV incorporates the properties of the HRRP samples themselves by kernel attention module. Finally, we present a lightweight SI-based network. The experimental results based on the measured HRRPs validate the effectiveness of proposed method across various extended recognition tasks, underscoring the potential of fusing SI via parameterized convolution in advancing target recognition systems.
BPFAaC: A radar intelligent perception algorithm in 6G integrated sensing and communications scenarios
2024, Expert Systems with Applications
In the era of scarce spectrum resources, the integration of communication and sensing will become a trend. The combination of radar and communication enables the synchronous transmission of integrated waveform information and radar detection function in the system, eliminating spectrum waste and maximizing equipment efficiency. In this article, we propose a radar sensing algorithm, Beta process factor analysis augmented classifier (BPFAaC), for the 6G integrated sensing and communications (ISAC) scenes. The integrated waveform's adaptive adjustment is guided by the detection sensing algorithm's outcomes. The BPFAaC model, which we propose, takes advantage of the later variable support vector machine (LVSVM) to learn a discriminative subspace with max-margin constraint. To enhance the performance of the prediction, it jointly learns the feature space and the augmented variable classifier in one framework. In this way, BPFAaC combines the advantages of Bayesian model to capture the latent feature of data, Bayesian nonparametrics to determine the number of factors, and LVSVM to induct label information. Moreover, it is natural for BPFAaC to handle outlier rejection problem, which benefits from the data description ability of BPFA. The proposed approach is validated through experiments with the measured HRRP datasets. The experimental results demonstrate that the proposed BPFAaC model outperforms baseline models and achieves more than 91.5% prediction accuracy.
TEFISTA-Net: A learnable method for high-resolution range profile reconstruction with low-frequency ultra-wideband radar
2024, Signal Processing
In low-frequency ultra-wideband (LFW) radar, the reconstruction of high-resolution range profile (HRRP) is an important task. The state-of-the-art compressive sensing (CS) method for HRRP reconstruction based on geometrical theory of diffraction (GTD) requires manual algorithm parameter tuning and is computationally expensive. In this paper, we propose a deep learning-based CS method for LFW radar HRRP reconstruction. We design a neural network architecture, i.e., target enhancement-based FISTA-Net (TEFISTA-Net), by unrolling the fast iterative shrinkage thresholding algorithm (FISTA). A new loss function based on the target-to-background ratio (TBR) is introduced for network training with the target enhancement capability in low signal-to-noise ratio (SNR) scenarios. The algorithm parameters in the traditional CS methods are substituted by network parameters learned from training data, getting rid of manual parameter tuning. In addition, simple convolution operations in our new method lead to lower computational complexity compared with existing methods. Experimental results based on diverse data show that the proposed method leads to higher computational efficiency with similar or better performance in comparison with existing state-of-the-art methods.
Quadruplet depth-wise separable fusion convolution neural network for ballistic target recognition with limited samples
2024, Expert Systems with Applications
Radar high-resolution range profile (HRRP) plays a crucial role in ballistic target recognition due to its simplicity and fast computation. Convolutional neural networks (CNNs), a popular deep learning approach, are widely used in radar automatic target recognition (RATR) based on HRRP, thanks to their excellent feature extraction capabilities. However, current research mainly focuses on integrating branch networks like recurrent neural networks and attention mechanisms into vanilla CNNs to enhance feature extraction. Limited research exists on reducing the computational complexity of vanilla CNNs when applied to HRRP-based RATR, as well as target recognition problems with limited samples. To address these issues, this paper proposes a lightweight depth-wise separable fusion CNN (DSFCNN) for ballistic target HRRP recognition. The DSFCNN reduces the computational complexity of vanilla CNNs while improving recognition accuracy. Furthermore, we introduce the sample-level fitting and class-level distinguishability (SFCD) loss for limited-sample ballistic target HRRP recognition. The SFCD loss ensures better fitting performance of all samples, as well as higher intra-class compactness and inter-class separability, enhancing the distinguishability of embedded features in limited-sample conditions. During the training phase, the SFCD loss is equally applied to four DSFCNNs with shared structures and parameters, resulting in a model known as Quadruplet DSFCNN (QDSFCNN). However, during the testing phase, only one DSFCNN is utilized. Additionally, we propose the parameter quantity shifting-fitting performance (PQS-FP) coordinate system to elucidate the performance disparities between DSFCNN and vanilla CNNs in our experiments. PQS-FP offers inspirational insights into the relations among parameter quantity, fitting performance, and recognition performance. Extensive experimental results demonstrate the robustness and effectiveness of QDSFCNN for ballistic target recognition in various limited-sample scenarios.
Target-attentional CNN for Radar Automatic Target Recognition with HRRP
2022, Signal Processing
Citation Excerpt :
Liu et al. [6] exploit positive alpha-stable distribution to characterize the distribution of HRRP data from sea clutter, and then obtain the more precise detection threshold for constant false alarm rate (CFAR)-based ship detection, thus leading to a better detection performance. In [9,15,22,28], different models (including local factor analysis, multi-task factor analysis, class factorized complex variational auto-encoder and target-aware recurrent attentional network) are developed to extract more information that can depict essential characteristics of the target from airplane HRRP data, and ultimately realize the recognition of airplane types, such as the recognition of Yark-42 and Cessna Citation S/II both belonging to jet plane. In this paper, we intend to achieve the recognition of target types with HRRP, referred to as the radar HRRP target recognition for convenience.
In this paper, a target-attentional convolutional neural network (TACNN) combining the convolutional neural network (CNN) and attention mechanism is proposed for radar high-resolution range profile (HRRP) target recognition. The TACNN takes one-dimensional CNN (1-D CNN) as the feature extractor and has the capability to excavate abundant local structural features of data. However, the HRRP contains non-target areas, where the information is useless or even unfavorable. Furthermore, different parts of HRRP target regions should have differences in contribution to the recognition task. Therefore, it is an inadvisable approach that treats all local features alike and directly uses them for the subsequent target recognition, which is adopted by a lot of models, such as the conventional CNN. To tackle this problem, the TACNN introduces the attention mechanism on the basis of 1-D CNN. In detail, the constructed attention module adaptively assigns a weight to each local feature of HRRP so as to locate the target areas and meanwhile enhance the interest of model in valuable target information. Specially, the attention mechanism in TACNN is realized via a bidirectional gated recurrent unit (Bi-GRU) network, where the attention coefficients used for weighting up local features are generated with full consideration of sequential relationship among different regional features in HRRP. Therefore, the learned attention coefficients in our TACNN can better represent the importance of each local feature to the recognition task, ultimately beneficial for the discovery of target information with more discriminability. Experimental results on measured HRRP data show that the proposed model can get more effectiveness in target recognition than related methods.

View all citing articles on Scopus

View full text

Target-Aware Recurrent Attentional Network for Radar HRRP Target Recognition

Highlights

Abstract

Introduction

Section snippets

Problem formulation

Proposed model

Experimental results

Conclusions

Acknowledgment

Expert Syst. Appl.

Iet Radar Sonar Navig.

Pattern Recognit.

Cogn. Sci.

Speech Commun.

Neurocomputing

Large margin feature weighting method via linear programming

IEEE Trans. Knowl. Data Eng.

A new feature vector using selected bispectra for signal classification with application in radar target recognition

Signal Process. IEEE Trans.

Bayesian spatiotemporal multitask learning for radar HRRP target recognition

IEEE Trans. Signal Process.

Combinatorial discriminant analysis: supervised feature extraction that integrates global and local criteria

Identification of ground targets from sequential high-range-resolution radar signatures

IEEE Trans. Aerosp. Electron. Syst.

A noise-robust radar target classification method based on complex probabilistic principal component analysis

General Assembly and Scientific Symposium

Radar HRRP statistical recognition with local factor analysis by automatic bayesian ying-yang harmony learning

IEEE Trans. Signal Process.

Radar HRRP target recognition based on dynamic multi-task hidden markov model

Proceeding of Empirical Methods in Natural Language Processing (EMNLP)

Radar HRRP target recognition based on k-SVD algorithm

IEEE Cie International Conference on Radar

A two-distribution compounded statistical model for radar HRRP target recognition

IEEE Trans. Signal Process.

Multi-task hidden markov model for radar automatic target recognition

IEEE Cie International Conference on Radar

Radar automatic target recognition based on complex high-resolution range profiles

International Conference on Radar

Automatic target recognition in high resolution sar image based on backscattering model

Dianbo Kexue Xuebao/Chin. J. Radio Sci.

Multi-task hidden markov modeling of spectrogram feature from radar high-resolution range profiles

EURASIP J. Adv. Signal Process.

Robust statistical feature based aircraft identification

IEEE Trans. Aerosp. Electron. Syst.

Neural machine translation by jointly learning to align and translate

Int. Conf. Learn. Represent.