This paper investigates the mixed-timescale resource allocation problem in satellite-terrestrial integrated networks (STIN). Moreover, the multi-access edge computing (MEC) technology and millimeter wave (mmWave) with rich spectrum resource are merged into the STIN to improve the network performance. A network utility maximization problem characterized by the achievable rate and backhaul reduction is formulated under the constraints of the maximum caching capacity, transmission power of mmWave small-cell base stations (SBSs) and quality of service (QoS) for Internet of Things (IoT) devices, where the caching placement, power allocation and user-SBS association are jointly optimized. In order to tackle this mixed-integer nonlinear programming (MINLP) problem, we decompose the original problem into the long-term caching placement subproblem, and short-term power allocation and user-SBS association subproblems. Then, a multi-agent deep reinforcement learning (MADRL)-based independent proximal policy optimization (IPPO) algorithm is proposed to solve the short-term user-SBS association subproblem. Meanwhile, the linear programming (LP) is used to solve the long-term caching placement subproblem. Furthermore, we derive the closed-form solution of the short-term power allocation subproblem through the Karush-Kuhn-Tucker (KKT) conditions. Simulation results are carried out to validate the effectiveness and scalability of the proposed joint approach.

Affective Image Manipulation (AIM) aims to alter visual elements within an image to evoke specific emotional responses from viewers. However, existing AIM approaches rely on rigidone-to-onemappings between emotions and visual cues, making them ill-suited for the inherently subjective and diverse ways in which humans perceive and express emotion. To address this, we introduce a novel task setting termedDiverse AIM (D-AIM), aiming to generate multiple visually distinct yet emotionally consistent image edits from a single source image and target emotion. We proposeEmoAgent, the first multi-agent framework tailored specifically for D-AIM. EmoAgent explicitly decomposes the manipulation process into three specialized phases executed by collaborative agents: a Planning Agent that generates diverse emotional editing strategies, an Editing Agent that precisely executes these strategies, and a Critic Agent that iteratively refines the results to ensure emotional accuracy. This collaborative design empowers EmoAgent to modelone-to-manyemotion-to-visual mappings, enabling semantically diverse and emotionally faithful edits. Extensive quantitative and qualitative evaluations demonstrate that EmoAgent substantially outperforms state-of-the-art approaches in both emotional fidelity and semantic diversity, effectively generating multiple distinct visual edits that convey the same target emotion.

To address the challenges of Internet of Things (IoT) device diversity and media service heterogeneity in human and machine-type communications, a predominant approach in sixth-generation (6G) networks and beyond is to serve diversified IoT devices by differentiated services. In this paper, a satellite-terrestrial IoT framework with multi-access edge computing (MEC) is investigated for two types of heterogeneous services, data-intensive and computation-intensive service. In our proposed framework, MEC and millimeter wave (mmWave) communication are jointly considered to optimize data- and computation-intensive services, guaranteeing the rate, delay and energy requirements of diversified IoT devices. From the viewpoint of heterogeneous services, we formulate a joint resource allocation problem, in which quality of experience (QoE) of diversified IoT devices are recognized as system utility. Specifically, service offloading, power allocation and computation resource allocation are jointly considered. Since the optimized problem is nonconvex, necessary problem reformulations are conducted to transfer the original problem to convex problems. Furthermore, an alternating iterative method based on deep reinforcement learning (DRL) and CVX technique is adopted to obtain the sub-optimal solution with low computation complexity. Finally, extensive simulations are conducted with different system parameter configurations to verify the effectiveness of our proposed scheme.

Human values are core principles to determine what is right, desirable, and important for individuals and societies. The deep integration of large language models (LLMs) into human life has facilitated their remarkable performance in value understanding. Multimodal content is rich in value-laden information. The development of multimodal large language models (MLLMs) provides new perspectives for multimodal value understanding. However, MLLMs’ capacity for understanding basic human values in the video domain remains underexplored. To bridge this gap, our work focuses on evaluating their ability to understand specific human values embedded in videos. We assess 11 advanced MLLMs based on the VVALUES video dataset, using a four-module strategy designed to answer two questions: Do MLLMs understand human values, and Do their design and training elements impact the performance? Based on the evaluation results, we derive valuable findings across eight aspects that provide deeper insights into the value understanding of MLLMs and offer useful guides for future research in this field.

In this paper, we propose a semantic communication framework facilitated by multi-access edge computing (MEC)assisted satellite-air-ground integrated networks (SAGIN), which comprises of LEO satellites, unmanned aerial vehicles (UAVs), and macro-cell base stations (MBSs). Considering the limited wireless resources and diversified quality of service (QoS) requirements of semantic tasks, an optimization problem with the goal of minimizing system cost in terms of the task latency and energy consumption is formulated. In order to address the mixed-integer nonlinear programming (MINLP) problem, we propose an alternating optimization algorithm that tackles UAV deployment sub-problem with the successive convex approximation (SCA) method, task offloading, semantic compression, power allocation and computation resource allocation optimization with deep reinforcement learning (DRL)-based multi-agent proximal policy optimization (MAPPO) method. Simulation results demonstrate that our proposed algorithm outperformes other reinforcement learning algorithms, i.e., about 5.12%, 23.72% and 35.64% over PPO, DDPG, and A2C, respectively.

With the blossom of Internet of Things (IoT) services and applications, the big data volumes raised by the large number of IoT devices have posed great burden on the traditional terrestrial networks. Considering the advantages of multi-access edge computing (MEC) and reconfigurable intelligent surface (RIS), this paper investigates the MEC and RIS-assisted airground IoT networks, where the joint resource allocation problem is formulated to minimize the system energy consumption. To handle the proposed nonconvex optimization problem, we decompose it into three subproblems, i.e., the coded caching placement problem, the phase shift problem and the joint multi-user association and power allocation problem. Then, we propose a deep reinforcement learning (DRL)-based Proximal Policy Optimization (PPO) algorithm to solve the joint multiuser association and power allocation problem. Moreover, the CVX technique and exhaustive search method are respectively adopted to solve the coded caching placement problem and the phase shift problem. Simulation results demonstrate that our proposed algorithm outperforms the benchmark schemes.

With the rapid proliferation of intelligent internet of things (IoT) terminals, conventional terrestrial networks are increasingly strained by limited bandwidth, high latency, and constrained computation capabilities. Air-Ground integrated networks (AGIN) offer a promising solution through flexible deployment, including enhanced coverage and low latency. To enable intelligent services, we propose an air-ground multi-access edge computing (MEC) network for semantic communication. Within this network, users transmit compressed semantic task data to unmanned aerial vehicles (UAVs) and terrestrial small-cell base stations (SBSs) using a probabilistic semantic compression (PSC) technique. A joint resource optimization problem is developed to determine semantic compression, task allocation, computation resource allocation, power control, and user association, aiming to minimize the system cost. The optimization problem is then formulated as a Markov decision process (MDP) and solved by a proximal policy optimization (PPO) algorithm based on deep reinforcement learning (DRL). Simulation results show that the proposed method significantly outperforms three other DRL baselines, achieving a reduction up to 74.12% in overall latency-energy cost under diverse system configurations.

Multimodal content has become the mainstream communication medium in video sharing platforms such as TikTok and Twitter, containing rich values information. Understanding human values is of great significance to multimodal content analysis and can be applied to downstream tasks such as recommendation systems and value alignment. However, current studies on human values mainly focus on text and lack a multimodal perspective. In this work, we present a new multimodal human values video dataset called VVALUES, which contains 5,104 annotated videos along with their titles. The dataset is labeled with coarse-grained polarity tags of positive and neutral, and fine-grained tags including 13 classes of value vocabulary. Based on VVALUES, we further develop a pre-trained language model (PLM)-based multimodal method adopting a dual-transformer variant for value recognition, MMVR. Extensive experiments demonstrate that our method significantly improves the performance of understanding values in videos. To the best of our knowledge, we are the first to try to incorporate human values in video understanding , and VVALUES is the first multimodal video dataset for human values.

In this paper, a localization algorithm based on Bayesian tensor decomposition is proposed for frequency diverse array multiple-input multiple-output (FDA-MIMO) radar, which successfully achieves vehicle localization in intelligent transportation systems (ITSs). Considering that the FDA-MIMO radar array may suffer from unknown mutual coupling (UMC), the proposed algorithm first constructs selection matrices for elimination, and then models the received signals as a third-order complex-valued tensor. To reduce the computational complexity of tensor decomposition, real-valued processing and compression techniques are employed to transform the complex-valued tensor into a real-valued compressed one. Subsequently, the factor matrices are obtained by Bayesian tensor decomposition, from which the direction of arrival (DOA) and range of the vehicle are extracted. Finally, the vehicle location is determined through geometric relationships. Besides, the Cramér-Rao bounds (CRBs) for DOA and range are derived as a performance benchmark. The proposed algorithm is applicable to the manifolds of uniform linear arrays (ULAs) and uniform planar arrays (UPAs) with UMC. Unlike existing algorithms requiring prior knowledge of target numbers, the proposed algorithm realizes accurate vehicle localization under both known and unknown target numbers. Simulation results demonstrate the effectiveness and robustness of the proposed algorithm.

Recent advances in generative compression methods have demonstrated remarkable progress in enhancing the perceptual quality of compressed data, especially in scenarios with low bitrates. However, their efficacy and applicability to achieve extreme compression ratios (< 0.05 bpp) remain constrained. In this work, we propose a simple yet effective coding framework by introducing vector quantization (VQ)–based generative models into the image compression domain. The main insight is that the codebook learned by the VQGAN model yields a strong expressive capacity, facilitating efficient compression of continuous information in the latent space while maintaining reconstruction quality. Specifically, an image can be represented as VQ-indices by finding the nearest codeword, which can be encoded using lossless compression methods into bitstreams. We propose clustering a pre-trained large-scale codebook into smaller codebooks through the K-means algorithm, yielding variable bitrates and different levels of reconstruction quality within the coding framework. Furthermore, we introduce a transformer to predict lost indices and restore images in unstable environments. Extensive qualitative and quantitative experiments on various benchmark datasets demonstrate that the proposed framework outperforms state-of-the-art codecs in terms of perceptual quality-oriented metrics and human perception at extremely low bitrates (≤ 0.04 bpp). Remarkably, even with the loss of up to 20% of indices, the images can be effectively restored with minimal perceptual loss.

Although multi-beam multiplexing can be implemented merely by phase shifters with hybrid beamforming configured by the sub-connected subarray architecture since all the antennas share the same magnitude, they cannot be set to a predetermined value. To tackle this issue, a non-convex constraint to enforce the magnitudes to a fixed value is first introduced in this design and then an iterative method is employed to relax it into a convex one. In doing so, the weighting magnitudes of all antennas can be preset in advance according to given requirements and a more flexible solution with phase-only excitation is obtained for multi-beam multiplexing. Numerical results are presented to verify the effectiveness of the proposed approaches.

With the continuous growth in demand for intelligent services, future 6G networks need to support higher communication efficiency and efficient intelligent connections. Semantic communication technology integrates the meaning of information into data processing and transmission, making it a potential paradigm for 6G. Considering that current research on semantic communication systems mainly focuses on the extraction and encoding of semantic features, with less attention to the impact of channel coding during the communication transmission process on system performance. Therefore, based on the CNN-Transformer (CT) semantic feature extraction and encoding scheme, this paper introduces a polar encoder, designing the end-to-end semantic CT-Polar communication system model frame-work. Through simulation verification, the CT-Polar designed in this paper demonstrated excellent performance in signal recovery on different datasets.

Abstract Reconfigurable intelligent surface (RIS) can provide unprecedented spectral efficiency gains and excellent ability to manipulate electromagnetic waves. This article considered a RIS‐assisted multiuser multiple‐input multiple‐output (MIMO) downlink system, where the beamforming at the base station and RIS are jointly designed to maximize the sum‐rate. For the large dimension scenario and high‐rank beamforming matrix, the accurate deterministic approximations from random matrix theory are then utilized to simplify the RIS‐assisted MIMO systems. The asymptotical signal‐to‐interference‐plus‐noise ratio values obtained through random matrix theory is infinitely close to the theoretical limits calculated by accurately iteration. And the performance of the proposed algorithm computed via the sharing second‐order channel statistics matches that of the RIS algorithm with sharing full channel state information asymptotically. The deterministic approximations are instrumental to get improvement into the structure of the optimal beamforming and to reduce the implementation complexity in large‐scale MIMO system. Numerical simulations results are provided to evaluate and verify the accuracy of the asymptotic results obtained from the proposed algorithm in the finite system regime. With the complex operation process of large dimension matrix reducing to the deterministic approximations, a lower computational complexity can be obtained compared with other methods.

In the sixth-generation (6G) Internet of Everything (IoE) environment, integrated sensing and communication (ISAC) can improve the utilization of radio resources. The application of reconfigurable intelligent surface (RIS) and millimeter wave (mmWave) can improve the performance of the ISAC. In this article, we propose an ISAC algorithm for RIS-assisted multiuser mmWave multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) systems. The proposed ISAC algorithm can achieve simultaneous channel estimation, positioning and environment mapping. Considering the limited robustness of the traditional algorithms to noise, the proposed algorithm first uses a complex-valued depth residual convolution neural network (C-DRCNN)-assisted channel estimation algorithm by using the powerful computational power of deep learning. Further, the sparsity of the mmWave channel can be utilized by the parallel factor (PARAFAC) tensor decomposition for obtaining the factor matrices, which contain the channel parameters, such as direction-of-arrival (DOA), direction-of-departure (DOD), time delay (TD), and complex path gain. Finally, the multiuser positioning and environment mapping are realized according to the geometric relationship between the position and channel parameters. The simulation results demonstrate that the proposed algorithm achieves better channel estimation, multiuser positioning and environment mapping performance compared with the state-of-the-art algorithm. In addition, since the proposed algorithm integrates the deep neural network and tensor decomposition, it still has excellent ISAC performance even at low signal-to-noise ratio (SNR).

A wearable dual-band and dual-circularly polarized (DBDCP) antenna using a dodecagonal truncate pyramid-shaped artificial magnetic conductor (AMC) reflector for gain enhancement is proposed in this paper. Firstly, a compact deformed quadruple inverted-F antenna (QIFA) with meander-line-shaped radiation patches has been developed as the radiator. Then, to make this QIFA generate different circular polarization (CP) radiation characteristics in two frequency bands, two feeding networks are adopted for realizing lefthand and righthand CP properties simultaneously. Lastly, a novel AMC reflector is employed to improve antenna performance. The presented DBDCP antenna was fabricated to realize lefthand CP (LHCP) in the frequency band of 3.5-4.0 GHz (13.3%) and righthand CP (RHCP) in 5.4-5.9 GHz (8.8%). Due to installation of the AMC reflector, the gains of the antenna are enhanced by about 3-5.3 dB in the lower CP band. The achieved peak gains are about 11.9 dBic and 10.5 dBic at 3.5 GHz (LHCP) and 5.8 GHz (RHCP), respectively. Meanwhile, the antenna’s specific absorption rate (SAR) has been greatly reduced, which meets well the IEEE wearable device standards. It is found that the proposed DBDCP antenna is a promising candidate for the applications of 5G, industrial scientific medical (ISM), WLAN (5.8-GHz), and WiMAX (3.5-GHz) systems in industrial IoT scenarios.

In order to support colourful multimedia services with strict quality-of-service (QoS) requirements of user equipments (UEs), the space-air-ground integrated networks (SAGIN) can be taken as a promising approach to enhance network capacity. Among them, millimeter wave (mmWave) and edge caching promise to significantly improve the SAGIN performance due to the advantage in rich bandwidth resource and low latency, respectively. In this paper, we investigate the joint caching and resource allocation for multimedia services in SAGIN, where multimedia content requests can be simultaneously served by multiple access points (APs). Considering the delay-constraint of multimedia services, we then formulate a mixed-integer non-linear programming (MINLP) problem aiming at minimizing the service delay, which involves jointly optimizing coded caching (CC), power allocation (PA) and UEs-to-APs association (UA). We propose to find the optimal solution by employing an alternating iteration optimization framework. The optimal CC and PA problems are firstly addressed by utilizing convex optimization technology. Then, two many-to-many swap matching algorithms are developed to slove the UA subproblem effectively. Numerical results demonstrate that our proposed algorithms can substantially reduce the service delay over other benchmarks.

The accelerated proliferation of visual content and the rapid development of machine vision technologies bring significant challenges in delivering visual data on a gigantic scale, which shall be effectively represented to satisfy both human and machine requirements. In this work, we investigate how hierarchical representations derived from the advanced generative prior facilitate constructing an efficient scalable coding paradigm for human-machine collaborative vision. Our key insight is that by exploiting the StyleGAN prior, we can learn three-layered representations encoding hierarchical semantics, which are elaborately designed into the basic, middle, and enhanced layers, supporting machine intelligence and human visual perception in a progressive fashion. With the aim of achieving efficient compression, we propose the layer-wise scalable entropy transformer to reduce the redundancy between layers. Based on the multi-task scalable rate-distortion objective, the proposed scheme is jointly optimized to achieve optimal machine analysis performance, human perception experience, and compression ratio. We validate the proposed paradigm's feasibility in face image compression. Extensive qualitative and quantitative experimental results demonstrate the superiority of the proposed paradigm over the latest compression standard Versatile Video Coding (VVC) in terms of both machine analysis as well as human perception at extremely low bitrates (< 0.01 bpp), offering new insights for human-machine collaborative compression.

In this paper, we propose a novel vehicle positioning method for commodity multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) WIFI systems in indoor parking lots. To address the limitation of the small number of WIFI antennas, the proposed method first utilizes signal model rearrangement techniques, in which the abundant carrier frequency resources of WIFI are expanded into space resources. Then a rearranged off-grid sparse Bayesian learning (ROG-SBL) algorithm is developed for parameters estimation to achieve vehicle positioning. Specifically, by resorting to the Bayesian inference and Newton method, the position-related parameters are estimated iteratively by fitting the channel state information (CSI) measurement model, and thus the vehicle positioning is realized according to the geometric relationship. Moreover, we derive the Cramér-Rao bound (CRB) as a performance reference for the proposed algorithm. Compared with the existing algorithms, the proposed one improves the positioning performance of the vehicle with fewer carrier numbers and has more stable performance. Simulation results show that the performance curves of the proposed algorithm for parameters estimation are close to the corresponding CRBs, and the proposed algorithm can cope with more challenging cases when the line-of-sight (LOS) path does not exist.

This paper proposes an effective algorithm for gain-phase error (GPE) and angle estimation in multiple-input multiple-output (MIMO) radar. First, the received signals are constructed into a third-order tensor model containing information such as GPE in transmitting arrays (Tx) and receiving arrays (Rx), directions-of-departures (DODs) and directions-of-arrivals (DOAs). Then, a tensor-based two-stage estimation algorithm to estimate GPE and angles is developed. In the first stage, GPE and angle information are obtained by the least squares Khatri-Rao factorization (LS-KRF) separation. In the second stage, the GPE in Tx and Rx is estimated by a two-step GPE estimation scheme, while the DODs and DOAs are estimated by a non-iterative angle estimation scheme based on the spatial smoothing preprocessing. Simulation results show the superiority of the proposed GPE and angle estimation algorithm.

Abstract In filter bank multi‐carrier (FBMC) with offset quadrature amplitude modulation (OQAM) systems, the major limitation of FBMC‐OQAM signaling is the high peak‐to‐average power ratio (PAPR). High PAPR results in increased hardware complexity and is also prone to non‐linear effects of high‐power amplifiers (HPA). The existing HPA energy‐efficient optimization strategies are constrained severely by the scarcity and heterogeneity of signal distribution in the degree of freedom. This paper theorizes the optimal FBMC‐OQAM signal distribution, which has a global and steady feature and evaluates the maximal HPA efficiency‐suited multi‐carrier modulation (MCM) signal. This paper first proposes that the Binary Particle Swarm Optimization based Partial Transmit Sequence (BPSO‐PTS) with the Input‐Back Off Modulation Error Ratio (IBO‐MER) scheme traces the signal distribution for maximal HPA efficiency to save supply‐side energy. This paper further proposes Continuous‐Unconstrained Particle Swarm Optimization based PTS(CUPSO‐PTS) with IBO‐MER to obtain the theoretical boundaries and drastically accelerate convergence in the continuous‐unconstrained searching space. To decrease the computational complexity of MER calculation, this paper proposes the CUPSO with IBO Signal‐to‐Distortion Ratio (IBO‐SDR), enabling it to meet more practical applications. The simulation results show that the proposed scheme outperforms conventional IBO and HPA efficiency schemes.

In film making, shot has a profound influence on how the movie content is delivered and how the audiences are echoed, where different emotions and contents can be delivered through well-designed camera movements or shot editing. Therefore, in pursuit of a high-level understanding of long videos, accurate shot detection from untrimmed videos should be considered as the first and the most fundamental step. Existing approaches address this problem based on the visual difference and content transitions between consecutive frames, while ignoring intrinsic shot attributes, viz., camera movements, scales and viewing angles, which essentially reveals how each shot is created. In this work, we propose a new learning framework (SCTSNet) for shot boundary detection by jointly recognizing the attributes and composition of shots in videos. To facilitate the analysis of shots and the evaluation of shot detection models, we collect a large-scale shot boundary dataset MovieShots2, which contains 15K shots from 282 movie clips. It is richly annotated with the temporal boundary between consecutive shots and its shot attributes, including camera movements, scales and viewing angles, which are the three most distinct shot attributes. Our experiments show that the joint learning framework can significantly boost the boundary detection performance, surpassing the previous scores by a large margin. SCTSNet improves shot boundary detection AP from 0.65 to 0.77, pushing the performance to a new level.

Abstract With the intensive deployment of mobile networks and the vigorous development of new multimedia services, video has gradually become the mainstream of cultural consumption. The contradiction between the proliferation of video data services and the scarcity of spectrum resources has brought great challenges to the current network resource allocation. Non‐orthogonal multiple access (NOMA) can be used to solve this problem by signal superposition and spectrum multiplexing to improve system access capability. As a new type of joint optimization design of transmitter and receiver side, PDMA has high research value. In this paper, a framework of PDMA video transmission system based on H.264 video compression coding (HVC‐PDMA) is proposed. Poly complementary sequence (PCS) spread spectrum coding is performed on the transmission codebook in order to improve the transmission accuracy. Meanwhile, a low complexity serial sphere compensated Max‐log MPA (SSCM‐MPA) algorithm is proposed to reduce the complexity of the multi‐user detection algorithm. Simulation results show that the PCS spread spectrum can improve system throughput and peak signal‐to‐noise ratio (PSNR) while reducing bit error rate (BER). SSCM‐MPA algorithm can greatly reduce the complexity and improve the transmission efficiency.

The rapid development of computer science and technology represented by artificial intelligence is fundamentally affecting man's lifestyle. Big data is pushing forward the development of artificial intelligence to unprecedented levels. The typical application, as one small element of the big data, refers to the knowledge engineering which sets the knowledge map as the core. A knowledge map is the knowledge base of the Semantic Web. Semantic web stands for a network describing things in the way that is understood by computers. The purpose of it is to enable computers to understand data and build a knowledge base of computers. By modeling the things that can be described, filling the attributes of things and expanding the connection with other things, the knowledge base can be built. This paper aims to propose a semantic information metadata model based on the ontology of traditional Chinese medicine and construct an ontology database with metadata model as the knowledge description method to realize keyword search by using a combination of ontology and Chinese medical knowledge data. Finally, achieving ontology database construction and visualization of ontologies.

In this letter, we present a dynamic adaptive multi-tree search (DAMS) algorithm which is an improvement of a multi-tree search algorithm (AMS). In order to reduce collisions, the multi-tree search algorithm adjusts the search tree through collision factor (CF). The proposed algorithm puts forward a new CF critical value and stipulates that the length of EPC which tags replied dynamically. The simulation results show that the DAMS search timeslots decrease about 20% than AMS, meanwhile DAMS reduces 70% transmission data than AMS.