In this paper, a study on bearing-only localization for wideband off-grid sources is presented by employing distributed sensor array networks. First, the wideband off-grid signal models based on distributed sensor array networks for both far-field and near-field scenarios are established with coarse grids employed for complexity reduction, where focusing algorithm is applied as a preprocessing step to facilitate multiple-frequency signal processing. Then, both one-step and two-step group-sparsity-based (GS-based) estimation methods are proposed for the considered two scenarios, where the on-grid information and the off-grid bias are recovered jointly (one-step methods) or alternatively (two-step methods). The proposed methods are capable of handling both overdetermined and underdetermined cases, leveraging information collected from all distributed receivers. To further alleviate off-grid approximation and focusing errors, a dynamic-dictionary-based iterative refocused grid-refining strategy is introduced, and the Cramér-Rao Bounds (CRBs) for uncorrelated source localization are also derived, which exists in the underdetermined case with more sources than subarray sensors. Simulation results demonstrate that improved performance can be achieved by our proposed off-grid solutions compared to existing methods.

In this paper, we propose variable window size (VWS) spatial smoothing techniques to improve coarray-based direction of arrival (DOA) estimation for sparse linear arrays (SLAs). The introduced approach exploits adjustable spatial smoothing window sizes, which yield significant gains in estimation accuracy for sparse arrays. Intuitively, the window size governs a fundamental bias-variance trade-off in the coarray domain: reducing it increases the number of averaged over lapping subarrays and strengthens the signal/noise subspace separation, but shortens the effective aperture or number of virtual coarray degrees of freedom. We formalize this trade off, establish an admissible compression range that preserves the signal/noise subspaces, and show that within this range smaller windows enhance the Subspace SNR. Specifically, we develop VWS Coarray Multiple Signal Classification (VWS-CA-MUSIC), VWS Coarray Root Multiple Signal Classification (VWS-CA rMUSIC) DOA estimation algorithms and their corresponding generalized versions. These algorithms are designed for fully and partially calibrated SLA geometries. An analysis of the proposed approach in terms of subspace properties is carried out along with an assessment of the computational complexity of the proposed algorithms. The proposed VWS algorithms offer three key advantages over fixed-window coarray methods: (i) improved root-mean-square error (RMSE) by incorporating extra unperturbed covariance components into the spatial smoothing average; (ii) enhanced Subspace SNR, which strengthens the separation between signal and noise subspaces; and (iii) a tunable compression parameter that allows designers to balance aperture and averaging according to the operating conditions. Numerical results assess the performance of the proposed techniques against competing approaches available in the literature.

Spectral reconstruction (SR) aims to learn the inverse mapping from RGB images to hyperspectral images. However, this remains a challenging ill-posed problem due to the severe information gap and overlapping spectral response ranges among RGB channels. To alleviate the representation-level entanglement caused by this inherent overlap, we propose a new Spectral Dual Matching Transformer (SDMT) based on a channel-decoupled representation and cross-channel collaborative fusion framework. Instead of processing the mixed RGB input monolithically, our framework extracts channel-specific features independently to prevent premature spectral crosstalk. Specifically, SDMT incorporates a Spectral Dual Matching self-Attention (SDMA) mechanism consisting of a sparse branch and a dense branch. The sparse branch utilizes Top-$k$ filtering to dynamically match highly correlated spectral bands, while the dense branch preserves long-range global correlations. Their features are adaptively fused to enhance key spectral details while maintaining overall background information. Furthermore, a Cross-channel Attention Fusion Module (CAFM) is developed to integrate the intermediate hyperspectral cubes from the decoupled channels. By executing cross-attention calculations among channels, CAFM effectively captures complementary information and rich inter-channel dependencies. Extensive experiments on the NTIRE2020, CAVE and TG1HRSSC benchmarks demonstrate that our proposed method is superior to the existing advanced methods.

Signals of opportunity (SoO) enable emission-free passive sensing, but low Earth orbit (LEO) satellite illumination with unmanned aerial vehicle (UAV) array receivers exhibits rapid geometry variation. As a result, the received phase evolves in a space-time coupled manner, and the array snapshots become nonstationary even within one coherent processing interval (CPI), degrading conventional stationary-snapshot direction-of-arrival (DOA) estimators. This paper proposes a decomposition-based sparse reconstruction with successive interference cancellation (D-SR-SIC) framework for dynamic DOA estimation in LEO SoO UAV passive sensing. The proposed estimator leverages a sparse-reconstruction signal model and is implemented via a computationally efficient decomposition-based search-and-cancel procedure. A short-CPI parameterized space-time phase model captures the common motion-induced phase history and the time-varying steering drift; the coupled multi-parameter estimation is decomposed into two low-dimensional correlation searches followed by least-squares amplitude estimation and multi-target peeling. Optional local refinement and multi-beam pre-screening improve robustness to off-grid mismatch, near-far interference, and wide field-of-view operation. Simulations show that the proposed method achieves about 0.11 degrees DOA root-mean-square error (RMSE) at -20 dB signal-to-noise ratio (SNR) in a representative highly dynamic setting.

Channel estimation is a pivotal challenge in reconfigurable-intelligent-surface-assisted (RIS-assisted) wireless communications. Despite the existence of numerous channel estimation methods, there remains a significant gap in addressing channel estimation problems in multi-user, multi-RIS, and multipath scenarios. In this paper, a two-stage channel refinement scheme based on both data-level and feature-level refinement of the channel is proposed. The first-stage refinement achieves channel estimation under low-bit RIS phase shifting conditions through joint design of the RIS phase shift matrices and the transmitted user pilot signals, thereby preliminarily enhancing the channel estimation accuracy. Subsequently, the second-stage further refines the estimation results via feature-level channel decomposition and reconstruction, effectively addressing the challenges posed by multi-user, multi-RIS, and multipath channels with further performance improvement. Moreover, Cramér-Rao lower bound (CRLB) for channel estimation is derived. Theoretical analysis and simulation results verify the effectiveness of the proposed scheme.

A two-stage cascaded direction of arrival (DOA) estimation framework is proposed for parameter estimation under non-ideal array conditions. The framework consists of two modules: contrastive reconstruction network (CRN) and DOA estimation network (DEN). To enhance robustness against array imperfections, CRN leverages self-supervised contrastive learning (SSCL) to recover covariance matrix, which is then fed into the DEN to perform the DOA estimation task. To further improve generalization, an augmentation strategy using non-ideal data, such as amplitude-phase perturbations and random sensor masking, is introduced during data training, where a two-stage training strategy is adopted to separately optimize the parameters of CRN and DEN. Simulation results demonstrate that the proposed framework achieves higher estimation accuracy and stronger robustness in non-ideal array conditions.

Under the frequency-decomposed wideband signal model, a group-sparsity (GS)-based wideband sparse representation of array covariance vectors (GS-WSRACV) framework is proposed by fully exploiting information from multiple subbands. First, an unweighted GS-WSRACV method is presented, for which a hyperparameter-tuning process is required. By prewhitening the residual constraint term in the optimization problem of GS-WSRACV (unweighted), a modified GS-WSRACV method is then developed, with a closed-form expression of the hyperparameter derived. Compared to existing prior-information-free sparse reconstruction-based direction-of-arrival (DOA) estimation methods, the proposed GS-WSRACV (unweighted) achieves the better estimation performance with a reduced computational complexity in each optimization; however, a hyperparameter-tuning process is required and the cost function for hyperparameter selection via trial-and-error is impractical due to the unknown actual DOAs for assessment, and the hyperparameter differs significantly with respect to the input signal-to-noise ratio (received source signal powers are also unknown). As a result, the proposed modified GS-WSRACV has achieved the best performance, and it is a more effective solution in practice without time-consuming hyperparameter-tuning process.

Terahertz (THz) communication combined with ultra-massive multiple-input multiple-output (UM-MIMO) technology is promising for 6G wireless systems, where fast and precise direction-of-arrival (DOA) estimation is crucial for effective beamforming. However, finding DOAs in THz UM-MIMO systems faces significant challenges: while reducing hardware complexity, the hybrid analog-digital (HAD) architecture introduces inherent difficulties in spatial information acquisition the large-scale antenna array causes significant deviations in eigenvalue decomposition results; and conventional two-dimensional DOA estimation methods incur prohibitively high computational overhead, hindering fast and accurate realization. To address these challenges, we propose a hybrid dynamic subarray (HDS) architecture that strategically divides antenna elements into subarrays, ensuring phase differences between subarrays correlate exclusively with single-dimensional DOAs. Leveraging this architectural innovation, we develop two efficient algorithms for DOA estimation: a reduced-dimension MUSIC (RD-MUSIC) algorithm that enables fast processing by correcting large-scale array estimation bias, and an improved version that further accelerates estimation by exploiting THz channel sparsity to obtain initial closed-form solutions through specialized two-RF-chain configuration. Furthermore, we develop a theoretical framework through Cramér-Rao lower bound analysis, providing fundamental insights for different HDS configurations. Extensive simulations demonstrate that our solution achieves both superior estimation accuracy and computational efficiency, making it particularly suitable for practical THz UM-MIMO systems.

Sparse reconstruction techniques including off-grid processing are widely applied in direction-of-arrival (DOA) estimation. However, multi-dimensional grid support is required in multi-dimensional estimation scenarios, which is computationally expensive and difficult to implement. In this work, a three-dimensional (3D) DOA and polarization joint estimation problem is studied and a decoupled off-grid signal model is first established, which decomposes the 3D estimation problem into two separate processes: DOA estimation and polarization parameters estimation. Then, an off-grid algorithm is proposed to estimate DOA based on a dynamic dictionary strategy, in which the algorithm no longer uses the fixed grids but dynamically updates the grid points to achieve DOA estimation in the continuous domain. After that, a structure-constrained method is provided to further estimate the polarization parameters, which provides a closed-form solution without requiring additional estimation pairing. Simulation results demonstrate that the proposed algorithm outperforms existing methods under various challenging conditions, including scenarios with a small number of snapshots, low signal-to-noise ratios, and the presence of coherent signals.

Conventional radar array design mandates interelement spacing not exceeding half a wavelength (λ/2) to avoid spatial ambiguity, fundamentally limiting array aperture and angular resolution. This paper addresses the fundamental question: Can arbitrary electromagnetic vector sensor (EMVS) arrays achieve unambiguous reconfigurable intelligent surface (RIS)-aided localization when element spacing exceeds λ/2? We provide an affirmative answer by exploiting the multi-component structure of EMVS measurements and developing a synergistic estimation and optimization framework for non-line-of-sight (NLOS) bistatic multiple input multiple output (MIMO) radar. A third-order parallel factor (PARAFAC) model is constructed from EMVS observations, enabling natural separation of spatial, polarimetric, and propagation effects via the trilinear alternating least squares (TALS) algorithm. A novel phase-disambiguation procedure leverages rotational invariance across the six electromagnetic components of EMVSs to resolve 2π phase wrapping in arbitrary array geometries, allowing unambiguous joint estimation of two-dimensional (2-D) direction of departure (DOD), two-dimensional direction of arrival (DOA), and polarization parameters with automatic pairing. To support localization in NLOS environments and enhance estimation robustness, a reconfigurable intelligent surface (RIS) is incorporated and its phase shifts are optimized via semidefinite programming (SDP) relaxation to maximize received signal power, improving signal-to-noise ratio (SNR) and further suppressing spatial ambiguities through iterative refinement.

This paper proposes a geometrically constrained decentralized independent vector analysis (GC-Dec-IVA) method for distributed microphone arrays. Recently proposed Dec-IVA method enables source separation by exchanging only power-related statistics to exploit cross-array information. However, this initial attempt often provides negligible improvement over applying IVA locally at each array, mainly due to the potential permutation inconsistency among arrays and the strong cross-array dependency implied by its source model. To address these limitations, we incorporate direction-of-arrival (DOA) information to derive GC-Dec-IVA, which mitigates permutation mismatch across arrays and enhances source alignment. Furthermore, a new source model is introduced to weaken cross-array dependency, improving robustness against permutation inconsistency in noisy environments. Experiments show the proposed method improves both the separation performance and cross-array permutation consistency.

Reconfigurable intelligent surface (RIS) can provide new ideas and methods for interference management. Particularly, applying interference alignment (IA) to RIS-aided interference networks can eliminate the interference completely and improve the system capacity enormously. However, the existing schemes solve the RIS-aided IA design problem by using iterative algorithms, which have high computational complexity. In this paper, we study the active RIS-aided IA design in closed form, for the multi-user multiple-input multiple-output (MIMO) system, comprising one active RIS and multiple transceiver pairs with multiple antennas. We first propose a novel framework of joint IA precoding and active RIS beamforming, to achieve the concurrent transmission of multiple interference-free data streams. Then, under the framework, the RIS beamforming is designed to minimize the rank of the MIMO interference channel, by using a matrix rank-reduced approach. After designing the RIS beamforming, the IA precoding at transmitter is designed to eliminate the interference completely. Furthermore, the proposed scheme can achieve the optimum degrees of freedom (DoF). Simulation results reveal that the proposed scheme outperforms the benchmark by 300% in terms of system sum-rate.

A group-sparsity-based (GS-based) underdetermined wideband direction-of-arrival (DOA) estimation framework, which is free of hyperparameters, is proposed to resolve more sources than the sensor number. Based on the subband model via frequency decomposition, a direct wideband extension of the narrowband SPICE method is first presented, fusing subband results jointly. Then, in order to tackle the underdetermined DOA estimation problem with a uniform linear array (ULA), a GS-based wideband sparse iterative covariance estimation (GSWSPICE) framework is proposed, where information acquired by subbands of interest are exploited simultaneously. Specifically, considering two a priori information on source spectrum under the proposed framework, the wideband cost function is reformulated to yield two improved objective functions and constraint structures, respectively. It significantly reduces the estimation parameters while fully exploiting the subband model's enhanced degrees of freedoms. Accordingly, two methods, i.e., GS-WSPICE (${{\mathcal {P}}_{\rm{up}}}$) and GS-WSPICE (${{\mathcal {P}}_{\rm{uf}}}$), are introduced as effective underdetermined solutions, where sparse arrays are not needed. Proposed methods are proved to be convex and can be expressed in semidefinite programming (SDP) form. Then, an efficient Cyclic Solution is developed, while a dynamic-dictionary-based hybrid Cyclic and SDP Solution is proposed to further reduce computational complexity. Numerical and experimental results validate the effectiveness of proposed hyperparameter-free methods, where the time-consuming hyperparameter tuning process is not required any more compared to existing wideband methods.

Current multiband synthetic aperture radar (SAR) feature extraction methods often fail to adequately exploit intraband details and interband spectral relationships, limiting their classification performance. To address this, in this article, a novel band-enhanced tensor-based multiband SAR feature extraction method is proposed for land cover classification. It first constructs a fourth-order feature tensor by aggregating gray-level co-occurrence matrix features from spatial neighborhoods, effectively capturing multiband spatial-spectral characteristics. Subsequently, a band-enhanced tensor discriminant locality preserving projection (BTDLPP) method is proposed to project the high-dimensional feature tensor to a discriminative compact subspace. BTDLPP innovatively integrates manifold structure preservation with linear discriminant analysis by formulating an objective function that simultaneously maximizes the between-class scatter and minimizes the within-class scatter and the manifold structural distortion. For the band mode, diagonal dominance and correlation guided regularization terms are added to the optimization objective function to enhance band information through weighted fusion. This approach enables comprehensive extraction of both linear and nonlinear discriminative features while jointly exploring intra and interband features. Experiments on multiband SAR datasets in Wanning and Sheyang regions demonstrate that, compared to the conventional tensor subspace projection methods, the feature space extracted by the proposed method simultaneously exhibits better linear separability and the preservation of manifold structures, leading to higher land cover classification accuracy.

As an approach to improve the uniform degrees-of-freedom (uDOFs) of sparse linear arrays, array motion, combined with the array dilation method, has received increasing attention, which increase the maximum uDOFs by several times through multiple uniform shifts and synthesis from array motion. In this work, an novel array dilation scheme adapted to arbitrary asymmetrically defined r-th-order cumulants with multiple shifts is proposed. By increasing the cumulants order or using longer shift sequences, a much higher number of uDOFs is achieved. Simulation results demonstrate the effectiveness of the proposed array dilation scheme.

In this work, we consider the challenging gridless direction-of-arrival (DOA) estimation problem with arbitrary sparse arrays. By utilizing the feature relationship between the received array signals and the gridless sparse recovery method, a local-global feature extraction self-supervision network (LGFS-Net) is proposed to achieve gridless sparse recovery without labels. The proposed network consists of a local-global feature extraction module and a complex atomic norm minimization (ANM) self-supervision module. The feature extraction module models the relationship of array elements to recover the missing data corresponding to the holes of the virtual array by learning the local and global features between the received data of different array elements. The complex ANM self-supervision module is embedded in the ANM problem, which realizes gridless Toeplitz matrix recovery and ensures the complex data structure in each network layer by complex singular value decomposition (SVD), positive semidefinite constraints (PSD), and ANM loss. By training the network using the received data of an arbitrary sparse array and constructing a self-supervised loss function that does not require labels, the network exhibits excellent performance when there are holes in the virtual array, and its signal recovery and parameter estimation performance is similar to the case of uniform linear arrays and better than traditional gridless recovery methods, as demonstrated by computer simulations.

This paper investigates the performance limitations of conventional linearly constrained minimum variance (LCMV) beamformers implemented with circular microphone arrays. In particular, we show that imposing null constraints on interference directions can lead to a deviation of the mainlobe from the desired steering angle. A theoretical analysis is presented to characterize this deviation, and a closed-form expression for the deviation is derived to reveal the underlying low-frequency behavior of LCMV beamformers. To address this issue, we propose a mainlobe-controlled LCMV (MC-LCMV) beamformer. Simulation results demonstrate that the proposed method substantially improves spatial directivity and speech enhancement performance compared with conventional LCMV beamformers.

Moving target detection using active sonar systems is an effective approach for underwater safety monitoring, where complex reverberation from the seafloor, water surface, and water column can substantially degrade detection performance. Robust principal component analysis exploits the low-rank structure of reverberation and the sparsity of moving targets in multi-frame sonar images and has become a widely adopted technique for reverberation suppression. However, the performance of existing optimization-based suppression methods is constrained by the manual selection of the regularization parameter. To overcome this limitation, this paper formulates reverberation suppression for moving target detection within a Bayesian framework. The low-rank component is characterized using a hierarchical Gaussian Wishart distribution, while the non-low-rank components are represented by a Gaussian mixture model. This prior formulation more effectively captures the complex characteristics of reverberation interference. Variational Bayesian inference is applied to perform matrix decomposition, with a generalized approximate message passing algorithm introduced to avoid large-scale matrix inversions. After estimating the non-low-rank components, the moving target component is directly extracted. In addition, nonlinear accumulation is employed on the moving target component to derive the target trajectory. Experiments conducted on four different target datasets show that the proposed method achieves superior detection performance compared with other methods.

To realize three-dimensional (3-D) underdetermined parameter estimation for near-field (NF) sources by L-shaped nested arrays, two NF localization methods based on the exact spherical wavefront model are proposed, including the cumulant-based method and the covariance-based method, which utilizes the temporal-spatial domain cumulants and correlations of NF sources, respectively. The cumulant algorithm constructs virtual received data through delayed fourth-order cumulant calculations of the original received data, while the covariance algorithm involves delayed autocovariance and delayed cross-covariance calculations. Subsequently, spatial-spectrum based subspace method is applied to both cumulant-based and covariance-based 3-D NF localization, where the former requires a 3-D spectral search procedure and the latter involves a one-dimensional and a two-dimensional spatial spectrum estimator to obtain the 3-D parameters. Furthermore, the computational complexity, the maximum number of identifiable NF sources and performance analysis of two proposed algorithms are provided. Simulation results demonstrate that the two proposed algorithms can achieve underdetermined 3-D parameter estimation for NF sources without any matching processes procedure. While the covariance algorithm has lower computational complexity, the cumulant algorithm performs better in parameter estimation.

Diffusion- and flow-based generative models have achieved strong performance in speech enhancement, but they typically rely on an explicit time variable to specify the generation stage. Since the noisy speech condition provides a strong reference to the underlying clean speech, enhancement progress could be reflected by the relation between the current sample and the noisy speech. This raises the question of whether the enhancement stage needs to be specified to the network through an explicit time variable. To study this question, we introduce EquilSE, an equilibrium matching-based framework that models generative speech enhancement with a time-invariant conditional vector field and formulates inference as an optimization process. By removing explicit time input and introducing an equilibrium-inducing target, EquilSE achieves comparable performance to time-dependent baselines under few-step inference and provides improvements in the single-step regime. Furthermore, EquilSE remains competitive under cross-domain evaluation.