With the development of deep learning (DL), joint source-channel coding (JSCC) solutions for end-to-end transmission have gained a lot of attention. Adaptive deep JSCC schemes support dynamically adjusting the rate according to different channel conditions during transmission, enhancing robustness in dynamic wireless environment. However, most of the existing adaptive JSCC schemes only consider different channel conditions, ignoring the different feature importance in the image processing and transmission. The uniform compression of different features in the image may result in the compromise of critical image details, particularly in low signal-to-noise ratio (SNR) scenarios. To address the above issues, in this paper, a dual attention mechanism is introduced and an SNR-adaptive deep JSCC mechanism with a convolutional block attention module (CBAM) is proposed, in which matrix operations are applied to features in spatial and channel dimensions respectively. The proposedsolution concatenates the pooling feature with the SNR level and passes it sequentially through the channel attention network and spatial attention network to obtain the importance evaluation result. Experiments show that the proposed solution outperforms other baseline schemes in terms of peak SNR (PSNR) and structural similarity (SSIM), particularly in low SNR scenarios or when dealing with complex image content.

Unmanned aerial vehicles (UAVs) are increasingly applied in various mission scenarios for their versatility, scalability and cost-effectiveness. In UAV mission planning systems (UMPSs), an efficient mission planning strategy is essential to meet the requirements of UAV missions. However, rapidly changing environments and unforeseen threats pose challenges to UMPSs, making efficient mission planning difficult. To address these challenges, knowledge graph technology can be utilized to manage the complex relations and constraints among UAVs, missions, and environments. This paper investigates knowledge graph application in UMPSs, exploring its modeling, representation, and storage concepts and methodologies. Subsequently, the construction of a specialized knowledge graph for UMPS is detailed. Furthermore, the paper delves into knowledge reasoning within UMPSs, emphasizing its significance in timely updates in the dynamic environment. A graph neural network (GNN)-based approach is proposed for knowledge reasoning, leveraging GNNs to capture structural information and accurately predict missing entities or relations in the knowledge graph. For relation reasoning, path information is also incorporated to improve the accuracy of inference. To account for the temporal dynamics of the environment in UMPS, the influence of timestamps is captured through the attention mechanism. The effectiveness and applicability of the proposed knowledge reasoning method are verified via simulations.

This paper proposes a multi-access and multi-user semantic communication scheme based on semantic matching and intent deviation to address the increasing demand for wireless users and data. The scheme enables flexible management of long frames, allowing each unit of bandwidth to support a higher number of users. By leveraging semantic classification, different users can independently access the network through the transmission of long concatenated sequences without modifying the existing wireless communication architecture. To overcome the potential disadvantage of incomplete semantic database matching leading to semantic intent misunderstanding, the scheme proposes using intent deviation as an advantage. This allows different receivers to interpret the same semantic information differently, enabling multiplexing where one piece of information can serve multiple users with distinct purposes. Simulation results show that at a bit error rate (BER) of 0.1, it is possible to reduce the transmission by approximately 20 semantic basic units.

The new applications surge with the rapid evolution of the mobile communications. The explosive growth of the data traffic aroused by the new applications has posed great computing pressure on the local side. It is essential to innovate the computation offloading methods to alleviate the local computing burden and improve the offloading efficiency. Mobile edge computing (MEC) assisted by reflecting intelligent surfaces (RIS)-based unmanned aerial vehicle (UAV) is a promising method to assist the users in executing the computation tasks in proximity at low cost. In this paper, we propose an energy-efficient MEC system assisted by RIS-based UAV, where the UAV with RIS mounted relays the computation tasks to the MEC server. The energy efficiency maximization problem is formulated by jointly optimizing the UAV's trajectory, the transmission power of all users, and the phase shifts of the reflecting elements placed on the UAV. Considering that the optimization problem is non-convex, we propose a deep deterministic policy gradient (DDPG)-based algorithm. By combining the DDPG algorithm with the energy efficiency maximization problem, the optimization problem can be resolved. Finally, the numerical results are illustrated to show the performance of the system and the superiority compared with the benchmark schemes.

In response to the challenge posed by the complexity of the system and the difficulty in obtaining accurate channel state information (CSI) for millimeter wave communication assisted by intelligent reflecting surfaces (IRS), we propose a deep learning-based channel estimation scheme. The proposed scheme employs a hybrid active/passive IRS architecture, wherein the least square (LS) algorithm is initially utilized to acquire the channel estimate from the active elements. Subsequently, this estimation is interpolated to obtain a preliminary channel estimation and ultimately refined into an accurate estimate of the channel using the channel super-resolution convolutional neural network (Chan-SRCNN) deep learning network. The simulation results demonstrate that the proposed scheme surpasses LS, orthogonal matching pursuit (OMP), synchronous OMP (SOMP), and deep neural network (DNN) channel estimation algorithms in terms of normalized mean squared error (NMSE) performance, thereby validating the feasibility of the proposed approach.

The application of frequency selection surfaces (FSSs) is limited by large area, narrow bandwidth, low stopband inhibition and large ripple in the passband. A method for designing high-order wide band miniaturized-element frequency selective surface (MEFSS) with capacitance loading is introduced. The proposed structure is composed of multiply sub-wavelength interdigital capacitance layer, sub-wavelength inductive wire grids separated by dielectric substrates. A simple equivalent circuit model, composed of short transmission lines coupled together with shunt inductors and capacitors, is presented for this structure. Using the equivalent circuit model and electromagnetic (EM) model, an analytical synthesis procedure is developed that can be used to synthesize the MEFSS from its desired system-level performance indicators such as the center frequency of operation, bandwidth and stopband inhibition. Using this synthesis procedure, a prototype of the proposed MEFSS with a third-order bandpass response, center frequency of 2.75 GHz, fractional bandwidth of 8% is designed, fabricated, and measured. The measurement results confirm the theoretical predictions and the design procedure of the structure and demonstrate that the proposed MEFSS has a stable frequency response with respect to the angle of incidence of the EM wave in the ±30° range incidence, and the in-band return loss is greater than 18 dB, and the rejection in the stopband is greater than 25 dB at the frequency of 3.2 GHz.

The existing level set segmentation methods have drawbacks such as poor convergence, poor noise resistance, and long iteration times. In this paper, a fractional order distance regularized level set segmentation method with bias correction is proposed. This method firstly introduces fractional order distance regularized term to punish the deviation between the level set function (LSF) and the signed distance function. Secondly a series of covering template is constructed to calculate fractional derivative and its conjugate of image pixel. Thirdly introducing the offset correction term and fully using the local clustering property of image intensity, the local clustering criterion of image intensity is defined and integrated with the neighborhood center to obtain the global criterion of image segmentation. Finally, the fractional distance regularization, offset correction, and external energy constraints are combined, and the energy optimization segmentation method for noisy image is established by level set. Experimental results show that the proposed method can accurately segment the image, and effectively improve the efficiency and robustness of exiting state of the art level set related algorithms.

Caused by the environment clutter, the radar false alarm plots are unavoidable. Suppressing false alarm points has always been a key issue in Radar plots procession. In this paper, a radar false alarm plots elimination method based on multi-feature extraction and classification is proposed to effectively eliminate false alarm plots. Firstly, the density based spatial clustering of applications with noise (DBSCAN) algorithm is used to cluster the radar echo data processed by constant false-alarm rate (CFAR). The multi-features including the scale features, time domain features and transform domain features are extracted. Secondly, a feature evaluation method combining pearson correlation coefficient (PCC) and entropy weight method (EWM) is proposed to evaluate interrelation among features, effective feature combination sets are selected as inputs of the classifier. Finally, False alarm plots classified as clutters are eliminated. The experimental results show that proposed method can eliminate about 90% false alarm plots with less target loss rate.