With the rapid development of terrestrial NR networks and satellite networks, and the demand of global end-users for ubiquitous network connectivity, the integration of heaven and earth has become an important development direction in the future. This paper focuses on the current development status and development trend of NR network and high orbit satellite network, and analyzes the interference situation of NR system and satellite system in Adjacent Frequency Coexistence scenario under the future development trend of heaven and earth integration, as well as the dry interference protection strategy. By setting the system and interference simulation parameters of NR system and high orbit satellite system, by choosing appropriate antenna model and propagation model, and by using reasonable interference assessment method, the Adjacent Frequency Coexistence of the two systems is analyzed, and the possibility of the system coexistence and the interference protection strategy are derived finally.

With the rapid development of various industries, the demand for instant communication has surged, and interphones are widely used in various fields such as civil aviation, fire protection, public security, and property management. Due to the mature technology, competitive pricing, and large market stock of analog products, traditional analog FM interphones still dominate the market. In the situation where frequency resources are becoming increasingly scarce and electromagnetic environments are becoming increasingly complex, interference phenomena of FM radio signals occur frequently. The automatic identification of FM signals is particularly important for maintaining radio order. This paper focuses on the weak links in current FM signal monitoring work and develops a comprehensive, fast, high-precision, and simple FM signal automatic recognition system. It can recognize FM sub tone signals, FM intercom signals and FM broadcast signals in real time, and greatly improve monitoring capabilities and effectively maintain radio order.

Vehicular Edge Computing (VEC) seeks adept task offloading strategies to streamline resource allocation and elevate vehicular application performance. This paper delves into scrutinizing the effectiveness of an Actor-Critic-based Multi-Agent Deep Reinforcement Learning (MADRL) framework in orchestrating cost-minimized task offloading within VEC domains. Embracing the MADRL paradigm, this framework empowers adaptive decision-making for task allocation and execution in highly dynamic vehicular environments. By integrating deep reinforcement learning methodologies into multi-agent systems, the study endeavors to bolster the efficiency of task offloading strategies operating within VEC realms. The research emphasizes MADRL’s role in fostering agile and responsive task allocation paradigms within VEC, aiming to optimize resource utilization while curtailing costs associated with task execution. Through adaptive decision mechanisms, MADRL facilitates dynamic task allocation, enabling vehicles to seamlessly distribute computational tasks among onboard systems and proximal edge resources. This exploration illuminates the potential of MADRL-based strategies in refining task offloading efficiency, presenting a promising avenue for enhancing VEC’s computational resource management and overall performance.

The rapid progression in autonomous driving and vehicle networking technologies has catalyzed the emergence of advanced vehicle applications, aiming to augment traffic safety and the driving experience. This advancement, however, is challenged by the limited computational and storage capacities inherent in on-board vehicle systems. To address this, Vehicle Edge Computing (VEC) emerges as a pivotal solution, enhancing vehicular computational capabilities. In this context, we introduce a novel VEC task offloading model utilizing Deep Reinforcement Learning (DRL). This model leverages the otherwise idle computational resources available in vehicles to facilitate efficient edge computing offloading within heterogeneous networks. A key innovation in our approach is the integration of Reinforcement Learning (RL) with Deep Learning (DL), significantly improving the convergence efficiency of the system. We also introduce an enhanced Q-learning algorithm tailored to jointly address the task offloading and processing challenges in VEC. This algorithm is adept at making optimal offloading decisions, aiming to minimize the overall system cost, encompassing both latency and energy consumption. Through rigorous simulation, our results demonstrate that this improved Q-learning approach substantially reduces total system costs while concurrently improving the quality of service in VEC environments. Our study not only offers a robust framework for computation offloading in vehicular networks but also paves the way for future research in AI-driven vehicular technology optimizations.

The rapid evolution of wireless communication technologies has led to the emergence of 6G networks, promising unprecedented connectivity and data processing capabilities. Within the 6G landscape, vehicle edge computing has gained increasing importance due to its potential to enhance real-time applications and services for connected and autonomous vehicles. However, the seamless integration of vehicles into edge computing environments presents a significant challenge—balancing the trade-off between energy consumption and latency. This research paper explores the critical problem of achieving an equilibrium between energy efficiency and low-latency communication in vehicle edge computing within the context of 6G networks. We propose a novel approach that integrates energy-aware computing strategies and latency optimization techniques to strike the optimal balance between these two conflicting objectives. In the methodology section, we provide a comprehensive overview of our proposed approach, detailing how it measures and evaluates both energy consumption and latency in the context of vehicle-edge interactions. We leverage advanced algorithms and models to efficiently allocate computing resources, optimize task scheduling, and minimize data transmission delays, all while considering the energy constraints of the vehicular devices. To validate the effectiveness of our approach, extensive experiments were conducted in a simulated vehicular edge computing environment. The experimental results demonstrate a substantial reduction in energy consumption while maintaining low-latency communication, outperforming existing methods. We present these results in graphical form, offering clear visual evidence of the improvements achieved.

We present a blended conditional gradient approach for minimizing a smooth convex function over a polytope P, combining the Frank{–}Wolfe algorithm (also called conditional gradient) with gradient-based steps, different from away steps and pairwise steps, but still achieving linear convergence for strongly convex functions, along with good practical performance. Our approach retains all favorable properties of conditional gradient algorithms, notably avoidance of projections onto P and maintenance of iterates as sparse convex combinations of a limited number of extreme points of P. The algorithm is lazy, making use of inexpensive inexact solutions of the linear programming subproblem that characterizes the conditional gradient approach. It decreases measures of optimality (primal and dual gaps) rapidly, both in the number of iterations and in wall-clock time, outperforming even the lazy conditional gradient algorithms of Braun et al. 2017. We also present a streamlined version of the algorithm that applies when P is the probability simplex.

Connecting visual imagery with descriptive language is a challenge for computer vision and machine translation. To approach this problem, the authors propose a novel end‐to‐end model to generate descriptions for images. Some early works used convolutional neural network‐long‐short‐term memory (CNN‐LSTM) model to describe the image, where a CNN encodes the input image into feature vector and an LSTM decodes the feature vector into a description. Since two‐dimensional LSTM (2DLSTM) has property of translation invariance and can encode the relationships between regions in an image, they not only apply a CNN to extract global features of an image, but also use a multidirectional 2DLSTM to encode the feature maps extracted by CNN into structural local features. Their model is trained through maximising the likelihood of the target description sentence from the training dataset. Experiments on two challenging datasets show the accuracy of the model and the fluency of the language which is learned by their model. They compare bilingual evaluation understudy score and retrieval metric of their results with current state‐of‐the‐art scores and show the improvements on Flickr30k and MS COCO.

Continuous action recognition in video is more complicated compared with traditional isolated action recognition. Besides the high variability of postures and appearances of each action, the complex temporal dynamics of continuous action makes this problem challenging. In this study, the authors propose a hierarchical framework combining convolutional neural network (CNN) and hidden Markov model (HMM), which recognises and segments continuous actions simultaneously. The authors utilise the CNN's powerful capacity of learning high level features directly from raw data, and use it to extract effective and robust action features. The HMM is used to model the statistical dependences over adjacent sub‐actions and infer the action sequences. In order to combine the advantages of these two models, the hybrid architecture of CNN‐HMM is built. The Gaussian mixture model is replaced by CNN to model the emission distribution of HMM. The CNN‐HMM model is trained using embedded Viterbi algorithm, and the data used to train CNN are labelled by forced alignment. The authors test their method on two public action dataset Weizmann and KTH. Experimental results show that the authors’ method achieves improved recognition and segmentation accuracy compared with several other methods. The superior property of features learnt by CNN is also illustrated.

The authors consider the general problem of graph clustering. Graph clustering manipulates the graph‐based data structure and the entries of the solution vectors are only allowed to take non‐negative discrete values. Finding the optimal solution is NP‐hard, so relaxations are usually considered. Spectral clustering retains the orthogonality rigorously but ignores the non‐negativity and discreteness of the solution. Sym non‐negative matrix factorisation can retain the non‐negativity rigorously but it is hard to reach the orthogonality. In this study, they proposed a novel method named congruent approximate graph clustering (CAC), which can retain the non‐negativity rigorously and can reach the orthogonality properly by congruency approximation. Furthermore, the solution obtained by CAC is sparse, which is approximate with the ideal discrete solution. Experimental results on several real image benchmark datasets indicate that CAC achieves encouraging results compared with state‐of‐the‐art methods.

Human gait is of essential importance for its wide use in biometric person-identification applications. In this work, we introduce a novel spatio-temporal gait representation, Flow Histogram Energy Image (FHEI), to characterize distinctive motion information of individual gait. We first extract the Histograms of Optical Flow (HOF) descriptors of each silhouette image of gait sequence, and construct an FHEI by averaging all the HOF features of a full gait cycle. We also propose a novel approach to generate two different synthetic gait templates. Real and synthetic gait templates are then fused to enhance the recognition accuracy of FHEI. We also adopt the Non-negative Matrix Factorization (NMF) to learn a part-based representation of FHEI templates. Extensive experiments conducted on the USF HumanID gait database indicate that the proposed FHEI approach achieves superior or comparable performance in comparison with a number of competitive gait recognition algorithms.

The presence of duplicated regions in the image can be considered as a tell-tale sign for image forgery, which belongs to the research field of digital image forensics. In this paper, a blind forensics approach based on DWT (discrete wavelet transform) and SVD (singular value decomposition) is proposed to detect the specific artifact. Firstly, DWT is applied to the image, and SVD is used on fixed-size blocks of low-frequency component in wavelet sub-band to yield a reduced dimension representation. Then the SV vectors are then lexicographically sorted and duplicated image blocks will be close in the sorted list, and therefore will be compared during the detection steps. The experimental results demonstrate that the proposed approach can not only decrease computational complexity, but also localize the duplicated regions accurately even when the image was highly compressed or edge processed.