Underwater images serve as one of the most intuitive media for human perception of underwater environments. However, the limited bandwidth and susceptibility to noise in underwater acoustic (UWA) channels pose significant challenges for traditional image encoding and transmission methods, thereby hindering high‐quality image reconstruction. Semantic communication aims to shift from pixel‐level to semantic‐level transmission, enhancing both reliability and efficiency. This paper proposes a scene‐guided generative communication method for underwater images with semantic alignment. We decouple and transmit only the essential layout information of underwater scenes. This enables highly efficient compression. At the receiver, we employ a graph convolutional network (GCN) to correct layout distortions and a context‐aware diffusion model to generate realistic underwater images that preserve high semantic fidelity to the original. Experimental results demonstrate that, compared to pixel‐fidelity communication and other generative communication approaches, our method consistently achieves superior image reconstruction quality even under adverse channel conditions, such as extremely low signal‐to‐noise ratios (SNRs), and exhibits significant advantages in downstream tasks.

Compared to human speech and multimedia audio, the amplitude of lung respiratory sounds is extremely weak, with minimal differences between the respiratory sounds of various lung diseases. Traditional methods, such as Mel‐frequency cepstral coefficients (MFCCs) and the Fourier transform, struggle to accurately extract respiratory sound characteristics of different lung diseases. Therefore, this article innovatively employs an improved C 0 complexity method to analyze the subtle differences in respiratory sounds, identifying the optimal tuning factor value that captures more local information and better characterizes the nonlinear stochastic properties of respiratory sounds. Subsequently, the second component of the MFCC was extracted and fused with the improved C 0 complexity and short‐term energy to propose a novel Mel‐energy complexity (MEC) feature. Finally, the fuzzy C‐means (FCM) clustering method was employed to process the MEC features, detecting the periodic endpoints of respiratory sounds, specifically, the endpoints of respiratory sounds during inhalation and exhalation, thereby laying a technical foundation for further research on respiratory sound recognition and visual management. To evaluate performance, we established a clinically collected respiratory sound database containing respiratory sounds from 154 individuals across four categories, including asthma and chronic obstructive pulmonary disease (COPD), with sounds recorded from eight chest locations per participant. The proposed method achieved average accuracy and F1 scores of 71% and 72%, respectively, outperforming multiple respiratory sound endpoint detection (RSED) approaches while demonstrating high robustness, particularly in processing pulmonary crackles.

Securing communication between multiple users efficiently while there are too many potential eavesdroppers has become an important issue with the rise of the Internet of Things (IoTs). This paper extends on earlier research, moving from a single‐user and single‐eavesdropper scenario to a complex multiuser and multieavesdropper context, and incorporates an advanced physical layer security (PLS) technique for the first time. Using reconfigurable intelligent surfaces (RISs) enhances the strength and quality of signals for intended users, while those to the unintended users are suppressed. Real‐time control of the RIS phase shifts is enabled through a deep deterministic policy gradient (DDPG) algorithm and this control significantly changes the trade‐off between security and energy wastage. The simulation results demonstrate that the developed approach can scale up in densely populated urban centers, while increasing the bit error rate (BER) performance and the overall energy efficiency across different wireless mobile channels.

In this paper, a nonorthogonal multiple access (NOMA) network is considered to assist ground users’ communications with two reconfigurable intelligent surfaces (RISs) and one unmanned aerial vehicle (UAV) technologies. In this scheme, UAV acts as a relay while the first RIS aids ground‐to‐air (G2A) communication and the second RIS helps the air‐to‐ground (A2G) communication. Under the outage probability (OP) constraint, the problem is formulated to maximize the data rate of the ground users by optimizing the transmission powers of base station (BS) and UAV, power allocation coefficients to the users and phase shift of the reflecting elements (REs) of RISs. An iterative algorithm is proposed using artificial bee colony (ABC) method for solving the problem and improving the network performance. Proposed algorithm is validated by Monte Carlo simulations, and the impact of system parameters is investigated on the network performance. Numerical results clarify that by two RISs utilization, the performances of the network are approximately improved 25% and 40% in terms of the network data rate and total OP, respectively, in comparison with the bench mark algorithms.

Blind spot network (BSN) has gained increasing attention with its state‐of‐the‐art performance in self‐supervised image denoising. However, most existing BSN models are based on an unrealistic assumption of noise independence and use isotropic mask convolutions, which can lead to the loss of structural details in the denoised image. To address these limitations, we consider the spatially correlated noise and introduce directional adaptive downsampling and mask convolutions to the wavelet domain, resulting in a novel self‐supervised denoising method called wavelet‐adaptive BSN (WA‐BSN). Specifically, we design the direction‐adaptive pixel‐shuffle downsamplings (PDs) and apply them to the wavelet decomposition subbands, where the spatial‐correlated noise is eliminated and the inherent structure is well preserved in the wavelet domain. Then, based on the geometric direction of the wavelet subimages, we propose four shape‐adaptive mask convolutions of a smaller size for each wavelet subband in WA‐BSN. This enables adaptive pixel prediction within a structural neighborhood for each subband with reduced training time. Finally, total variation (TV) is added to the loss function to further preserve the edges. The results on public real‐world datasets demonstrate that our method significantly outperforms existing self‐supervised denoising methods and achieves great efficiency.

Eigenvalue detection is extensively utilized in numerous applications, including spectrum sensing in cognitive radio. However, the characteristics of the Tracy–Widom () distribution and its lack of accurate close‐form expression limit the detection performance based on the extreme eigenvalue. The paper introduces an eigenvector correlation (EVC) based detection. It extracts the eigenvalue of a weak component from the Marchenko–Pastur law (MP‐law) bulk by combining a priori known eigenvector‐correlated signal. The extracted eigenvalue then follows an easily analyzable and tractable Gaussian distribution according to the random matrix theory. The proposed method significantly improves sensing performance, which is theoretically analyzed and compared with traditional maximum eigenvalue‐based detection (MED) and trace‐based sensing. In addition, a cumulative EVC (CUM–EVC) is further developed for the multiple eigen‐component signal. Numerical results are presented to validate the theoretical analysis and demonstrate the reliability of the proposed detectors.

Wireless communication plays a crucial role in the automation process in the industrial environment. However, the open nature of wireless communication renders industrial wireless sensor networks susceptible to malicious attacks that impersonate authorized nodes. The heterogeneity of the wireless transmission channel, coupled with hardware and software limitations, further complicates the issue of secure authentication. This form of communication urgently requires a lightweight authentication technique characterized by low complexity and high security, as inadequately secure communication could jeopardize the evolution of industrial devices. These requirements are met through the introduction of physical layer authentication. This article proposes novel deep learning (DL) models designed to enhance physical layer authentication by autonomously learning from the frequency domain without relying on expert features. Experimental results demonstrate the effectiveness of the proposed models, showcasing a significant enhancement in authentication accuracy. Furthermore, the study explores the efficacy of various DL architecture settings and traditional machine learning approaches through a comprehensive comparative analysis.

High dynamic range omnidirectional image (HOI) can provide more real and immersive watching experience for viewers, thus has become an important presentation of virtual reality technology. However, both the system processing and the characteristics of HOI make the design of HOI quality metric (HOIQM) a challenging issue. In this work, considering the difference between whole field of view (FoV) and viewer‐selected viewport, distortion features from both global and local perspectives are extracted, and a blind HOIQM is proposed. Specifically, because different regions have different projections in SSP projection, we have constructed the optimal bivariate response pair in the equatorial region and bipolar region according to their projection direction, and parameters in the BGGD based‐spatial oriented correlation model are extracted as global statistical features. Meanwhile, combined with the visual perception for HOI, the key blocks are determined in equatorial region, and the local statistical characteristics of the key blocks are extracted by analyzing the distribution of multiscale structure information. Finally, the global and local features are regressed by SVR to obtain the final HOI quality. Experimental results on NBU‐HOID database demonstrate that the proposed quality metric is outperformed the existing representative quality metrics and is more consistent with human visual perception for HOI.

Compressive sensing (CS) is a technique that enables the recovery of sparse signals using fewer measurements than traditional sampling methods. To address the computational challenges of CS reconstruction, our objective is to develop an interpretable and concise neural network model for reconstructing natural images using CS. We achieve this by mapping one step of the iterative shrinkage thresholding algorithm (ISTA) to a deep network block, representing one iteration of ISTA. To enhance learning ability and incorporate structural diversity, we integrate aggregated residual transformations (ResNeXt) and squeeze‐and‐excitation mechanisms into the ISTA block. This block serves as a deep equilibrium layer connected to a semi‐tensor product network for convenient sampling and providing an initial reconstruction. The resulting model, called MsDC‐DEQ‐Net, exhibits competitive performance compared to state‐of‐the‐art network‐based methods. It significantly reduces storage requirements compared to deep unrolling methods, using only one iteration block instead of multiple iterations. Unlike deep unrolling models, MsDC‐DEQ‐Net can be iteratively used, gradually improving reconstruction accuracy while considering computation tradeoffs. Additionally, the model benefits from multiscale dilated convolutions, further enhancing performance.

Sample clustering techniques play a crucial role in the data‐driven state evaluation of electromechanical equipment, and selecting an appropriate similarity measurement method for sample sets helps improve the clustering performance. The Jaccard coefficient is a commonly employed indicator of similarity for scalar set‐type samples. In this paper, we propose an incremental clustering algorithm for matrix‐type samples by defining an improved Jaccard coefficient. First, a new binary relation is formulated to derive a relationship matrix between samples. Second, an undirected graph is given by using the relationship matrix, and an improved pruning operation is provided to simplify the graph by eliminating redundant edges. Then, a new relationship matrix is generated according to the modified graph, which enables the calculation of the improved Jaccard coefficient. By using the improved Jaccard coefficient, the improved incremental clustering algorithm updates cluster centers by selecting a particular sample to maximize the sum of similarities between the selected sample and other samples within the same cluster. Finally, the effectiveness of the proposed incremental clustering algorithm is demonstrated in fault diagnosis and remaining useful life estimation application scenarios, respectively. The experimental results indicate that the improved algorithm outperforms traditional clustering methods.