Streaming the intubation procedure performed with a videolaryngoscope (VL) to operating room monitors has been rendered possible through technical improvements. This study explored the impact of streaming McGRATH MAC+™ VL live intubations to operating room monitors on team dynamics, situational awareness and evaluated overall perceptions of the device in daily practice. This cross-sectional, observational study was conducted between February and November 2025 at the Lausanne university hospital (CHUV) in Switzerland. Operating room professionals with exposure to McGRATH MAC+™ streaming completed a paper-based questionnaire assessing perceived changes in awareness, participation, communication, and device performance. Retrospective within-subject comparisons (before vs after streaming ratings) were analysed. We enrolled 50 participants (37 anaesthesia, 13 operative). After the introduction of VL streaming, awareness of the intubation process was perceived to have improved by the anaesthesia staff (p < 0.001) and the operative staff (p = 0.005). The participants didn’t perceive a significant change in their participation to the intubation with VL streaming. Secondary outcomes highlighted facilitated intra-team communication within anaesthesia teams and modestly improved inter-team coordination. No meaningful changes were reported regarding perceived intubation time or complication rates. In addition, VL was considered highly effective for training and improved operator confidence. Streaming videolaryngoscope live intubations to operating room monitors resulted in higher retrospectively perceived situational awareness and the device was viewed as an educational asset. Selective use in difficult or teaching cases may offer a high-yield strategy for enhancing team coordination during airway management. clinicaltrials.gov : NCT06453525

AGNER: Agile governance-oriented unified named entity recognition for continual learning with diffusion adaptation.

Abstract— In today’s digital age, a face recognition system plays an important role in almost every field. Face recognition is one of the most commonly used biometric technologies. It can be applied for security, verification, identification, and offers many other benefits. Although its accuracy is comparatively lower than iris and fingerprint recognition, it is widely adopted because it is contactless and non-intrusive. Moreover, face recognition can be effectively utilized for recording attendance in schools, colleges, offices, and other institutions. This system is designed to develop a classroom attendance solution based on face recognition, as traditional manual attendance methods are time-consuming and difficult to manage. There is also a possibility of proxy attendance in manual systems, which increases the necessity for an automated solution. The proposed system operates in four stages: database generation, face detection, face identification, and attendance updating. The database is prepared using students’ images collected from the class. Face detection and identification are carried out using the Haar-Cascade classifier and the Local Binary Pattern Histogram (LBPH) algorithm, respectively. Faces are identified from real-time video streaming in the classroom. At the end of the session, the attendance report is sent to the respective faculty member via email. Keywords— Automated Face; Identification; Image-Based Face Detection,;Haar Feature-Based Cascade Classifier;Local Binary Pattern (LBP);Automated Attendance System

This study explores the application of predictive data science techniques to anticipate customer behavior in the mobile telecommunications sector, integrating insights from finance and marketing. Leveraging advanced predictive models, including a multitask learning approach, the research is supported by an interactive web interface featuring a home- page, a Power BI dashboard, and a prediction page. The aim is to transform raw historical customer data into action- able insights for marketing teams, enabling accurate forecasts of mobile internet package activation and customers’ potential future value. Findings indicate that customer satisfaction and perceived sustainable value significantly influence subscription and recharge decisions, thereby enhancing loyalty and revenue generation. By emphasizing the synergy be- tween business needs, technological tools, and methodological frameworks, this work offers an innovative combination of theoretical and empirical approaches to advance practices within the telecommunications industry. Future research directions include incorporating real-time data streams and developing automated marketing recommendations to further optimize strategic effectiveness.

In the context of increasingly complex data-driven decision-making, traditional analytical methods face significant challenges in handling heterogeneous data correlations and real-time computational responsiveness. By developing an Integrated Computational Business Analytics Framework (ICBAF), this research proposes a three-layer computational architecture comprising heterogeneous data representation, heuristic computational optimization, and online feedback loops, aiming to achieve a deep coupling of business logic and algorithmic reasoning. The findings indicate that by leveraging the topological representation capabilities of Graph Neural Networks (GNN) and the dynamic optimization mechanisms of real-time stream computing, this framework effectively resolves bottlenecks such as "data silos" and "decision lag," significantly enhancing decision precision in scenarios like intelligent marketing and smart operations. However, as the study primarily focuses on theoretical deduction and scenario-based simulations, it lacks rigorous empirical validation using large-scale, real-world industrial datasets. Consequently, its generalization capabilities under extreme noise and the cost-benefit ratio of computational resources require further verification in future research.

Introducing a workflow algorithm for adaptive DBS programming in Parkinson's disease.

The growing use of video surveillance systems requires reliable communication networks capable of supporting high-quality real-time streaming. This study evaluates the performance of a wireless network designed for CCTV transmission using the NS-3 simulation platform. The proposed network follows a star topology and consists of four communication links with different distances and bandwidth capacities to examine their behavior under varying traffic loads. This work is part of a broader research effort aimed at designing an efficient communication network to support road safety monitoring and traffic surveillance in mountainous urban environments. It also seeks to provide technical insights that may assist local authorities in developing reliable monitoring infrastructures and improving traffic management. The evaluation focuses on key Quality of Service (QoS) metrics, including throughput, delay, jitter, and packet loss. The results show that higher-capacity links maintain stable throughput and low delay under increased traffic load, while longer-distance links begin to experience congestion as the offered traffic approaches their capacity limits. These findings highlight the importance of link capacity in network performance and confirm that a properly designed star-based wireless infrastructure can effectively support traffic surveillance systems in geographically challenging environments.

The mining environment of molybdenum is highly complex and dynamic, with a large number of safety risks related to personnel operation and equipment operation. Traditional safety monitoring methods mainly rely on manual inspections, which are inefficient and unable to achieve real-time monitoring and precise early warning. To address the above challenges and enhance the safety management level of Chifeng Molybdenum Mine, this study designed and implemented an abnormal behavior recognition system based on deep learning. This system adopts an edge-cloud collaborative architecture, organically integrating the front end and the back end. The back end utilizes Node-RED for low-code process orchestration, integrates EdgeX to achieve unified access of heterogeneous devices, adopts EMQX to manage large-scale device communication, realizes real-time video stream distribution through MediaMTX, and introduces the optimized YOLOv8 model for AI reasoning and analysis. The system, based on computer vision technology, has achieved real-time monitoring of key safety elements such as face recognition, safety helmet detection, and distance monitoring for hot work operations. The experimental results show that the accuracy rate of abnormal behavior detection by the system in complex mining environments is significantly better than that of traditional monitoring methods.

Abstract paper presents WLASL (Word-level American Sign Language) as a dataset for building a voice to sign language translator that translates spoken English into ASL (American Sign Language) through video. The system integrates ASR (Automatic Speech Recognition) to transcribe (convert) real-time speech inputs into text. This transcription process identifies the respective keywords from an input sentence, and these keywords are then mapped to ASL glosses. Sign language videos of the respective ASL signs are streamed to the user through a web-based chatbot interface. The purpose of this digital communication tool is to provide greater accessibility for Deaf and hard of hearing individuals by improving access to various types of voice-to-sign communication systems. In developing the WLASL translator sistema we incorporate scalable architecture, real-time performance, and efficient retrieval of video data without the need for storing output videos on the server. This research informs the development of AI-based assistive technologies that have the potential to connect hearing and non-hearing communities. Keywords: Voice-to-Sign Translation, WLASL Dataset, American Sign Language (ASL), Automatic Speech Recognition (ASR), Assistive Technology, Real-Time Video Streaming, Accessibility.

This paper proposes a hybrid underwater drone communication system that combines underwater optical communication (UOC) and radio-frequency (RF) communication to support real-time video streaming in underwater environments. The system consists of a remotely operated vehicle (ROV) that transmits video to a surface gateway, which relays the video to onshore facilities through a 5G network. An outdoor experiment conducted in a maritime environment measured the received signal strength indicator (RSSI), signal-to-interference-plus-noise ratio (SINR), occupied bandwidth, and end-to-end (E2E) throughput at 700 MHz and 2600 MHz with video frame rates ranging from 10 to 60 fps. The results show that the 700 MHz frequency band provides higher RSSI and SINR, which support more reliable long-range communications, while the 2600 MHz frequency band provides lower RSSI and SINR but a larger bandwidth. The maximum E2E throughput achieved was 53.5 Mbps at 700 MHz and 58.64 Mbps at 2600 MHz. Increasing frame rates mainly affects throughput by reducing SINR. These results analyze the coverage–capacity trade-off and provide valuable insights for drone-assisted hybrid UOC-RF communication in underwater video streaming applications.

The proliferation of wireless sensor networks (WSNs) in the health care sector has enabled continuous, real-time monitoring of patients’ vital signs, enabling early diagnosis and prompt care. Still, the healthcare data classification techniques currently in use face critical challenges, including high computational cost and slow convergence, as well as limited scalability for large, real-time data streams. This research proposes the Optimized Healthcare Data Classification Model (OHD-Classifier) to streamline real-time monitoring systems by integrating Support Vector Machines (SVMs) with Stochastic Gradient Descent (SGD) (OHD-SVMSGD). As described, the OHD-Classifier advances SGD-trained SVMs, thereby accelerating convergence and improving efficiency. With this advancement, the OHD-Classifier improves classification accuracy for vital healthcare conditions, including irregular heart rates, critical- and low-variance hypertensive episodes, and other essential sign anomalies. The primary function of the model remains real-time data processing from WSNs, with provisions for instantaneous feedback to enable effective healthcare staff and decision-making. The solutions available today are hampered by slow model training due to the high dimensionality of the feature space, overfitting to WSNs, and unstable input data streams. The OHD-Classifier addresses the issues described through effective feature selection and the optimization of SVM parameters by SGD. This not only streamlines model training but also progresses the model’s generalization to novel data. Outcomes indicate the OHD-Classifier outperforms all competing models on accuracy, training speed, and adaptability. The proposed model has the accuracy as 98%, Precision as 97%, Recall as 97% and F1 Score as 99%. This research lays the groundwork for developing more efficient, scalable monitoring systems in the healthcare sector and for improving patient care and outcomes in fast-changing, low-resource environments.

Serious games are an important tool to overcome the low engagement and adherence to rehabilitation programs due to their repetitive nature and lack of positive reinforcement. Dynamic difficulty adjustment (DDA) systems can contribute by providing algorithms to adapt serious games, keeping players engaged and in a flow state. However, these systems are generally custom-made for specific purposes and goals, lacking the adaptability to be easily integrated into serious games. In response to this problem, we introduced the Flow Optimizer Framework (FOF), a game-agnostic DDA system developed for Unity. This framework facilitates the integration of DDA algorithms with serious games in Unity, enabling real-time monitoring and adaptation based on player state through data processing, rule-setting, and decision-making. First, we conducted a technical validation of the framework, assessing its performance in handling real-time data streams and its responsiveness to different scenarios. Following this validation, we evaluated its effectiveness in enhancing the flow state by conducting a usability study. Participants were presented with three different types of DDA paradigms implemented in FOF (Implicit, Explicit, and Subjective), each with different algorithms to adjust the game's difficulty. The results obtained showed that the implementation of a biofeedback paradigm using the player's heart rate was the one that increased game performance the most, and participants reported this condition as the most enjoyable and fitting to their skills. Overall, participants reported a high usability and a high presence experienced in the serious games implemented with FOF.

Telecom Rapid expansion of video conferencing, is a critical communication media, collaboration and Training to most of the verticals, especially after the outbreak is over. In this paper, the summary of the recent studies on the concepts of Evolution of video, ease of use and future direction in conference system is provided. There are early theories (Technology Acceptance Model (TAM) and Media Simultaneity Theory (MST) which explain adoption habits and user-involvement. User satisfaction fatigue Digital conferencing and QoE research has given latency and interface design and cognitive aspects as roles in User satisfaction. The latest developments introduce artificial intelligence, Real-time streaming (WebRTC), and 3D/volumetric streaming telepresence to offer a better degree of immersion and social presence, users with vision and hearing impairments do not enjoy the same degree of accessibility, scaling fatigue remain issues to allow a seamless experience. Some of the new trends that are increasingly being adopted are mixed reality, augmented reality based on communication and gesture support, and artificial intelligence. Debriefs will become the virtual world of interactivity and inclusivity that will transform the world, and they are justified. The information presented in this article gives a glimpse of the theoretical knowledge and Empirical information to comprehend the development of video conferencing by two domains: Immersive, intelligent, Human Centric communication ecosystems.

In real-world streaming recommender systems, user preferences evolve dynamically over time. Existing bandit-based methods treat time merely as a timestamp, neglecting its explicit relationship with user preferences and leading to suboptimal performance. Moreover, the online learning methods often suffer from inefficient exploration–exploitation during the early online phase. To address these issues, we propose HyperBandit+, a novel contextual bandit policy which integrates a time-aware hypernetwork to adapt to time-varying user preferences and employs a large language model-assisted warm-start mechanism (LLM Start) to enhance exploration–exploitation efficiency at the early online phase. Specifically, HyperBandit+ leverages a neural network that takes time features as input and generates parameters for estimating time-varying rewards by capturing the correlation between time and user preferences. Additionally, the LLM Start mechanism employs multi-step data augmentation to simulate realistic interaction data for effective offline learning, providing warm-start parameters for the bandit policy at the early online phase. To meet real-time streaming recommendation demands, we adopt low-rank factorization to reduce hypernetwork training complexity. Theoretically, we rigorously establish a sublinear regret upper bound that accounts for both the hypernetwork and the LLM warm-start mechanism. Extensive experiments on real-world datasets demonstrate that HyperBandit+ consistently outperforms state-of-the-art baselines in terms of accumulated rewards.

Multimedia data centres have grown as a result of the use of cloud services, IoT, and real-time streaming, but they are now more susceptible to sophisticated attacks. This study suggests a hybrid deep learning-based intrusion detection system (IDS) to identify a variety of attacks such as DoS/DDoS, Probe/Reconnaissance, R2L, U2R, and IoT-specific threats like ransomware and backdoors. We employ a supervised architecture to record the spatial-temporal patterns of known attacks, and an unsupervised Autoencoder to detect abnormalities. Using a majority voting method, the outputs of both models were combined to provide the final findings. A number of preparation stages, such as MinMax normalization, binary and one-hot label encoding, data forming to match deep learning models, and class balancing using strategies like SMOTE or under sampling, are required to make the methodology function in real-world condition. The best performance on CIC-IDS2018, with an accuracy of 99.1%, precision of 98.9%, recall of 99.3%, and F1-score of 99.1%, was achieved by the hybrid model when it was tested on three benchmark datasets: TON_IoT, CIC-IDS2018, and NSL-KDD. Additionally, strong results were shown by NSL-KDD (accuracy 98.2%) and TON_IoT (accuracy 97.3%), and the flexibility and adaptability of the suggested IDS were illustrated. Good accuracy, precision, recall, F1-score, and ROC-AUC are kept by the system while false positives and false negatives are reduced. These findings are showing that the suggested IDS is reliable, flexible, and capable of defending multimedia data centers against dynamic internet attacks.

This study investigates how people perceive and evaluate true and false news in their natural environment using a novel experience sampling methodology with real-time news streaming. The sample consisted of 110 participants who evaluated news headlines on their smartphones throughout the day for two weeks, receiving notifications when new content was published. The study employed a custom-developed server that captured RSS feeds from major news outlets. The server used AI (the Open AI “gpt-4-0125-preview” model) to generate modified versions of news stories on the fly, including misinformation variants. Participants evaluated news live under experimentally manipulated conditions that included time constraints for reading the news. They also provided information about their environmental context and individual characteristics. The results showed that false news items were generally rated as less accurate than true news, but this discernment decreased under time constraint. Higher digital literacy and greater satisfaction with the political system were associated with rating false news as less accurate, whereas higher dogmatism was linked to higher perceived accuracy of false news. Familiarity was also related to higher accuracy ratings for both true and false news, meaning participants rated both types of news as more accurate when they felt familiar with it, consistent with the illusory truth effect. Integrating experimental AI-guided, real-time news generation and streaming offers a novel and much-needed approach to studying misinformation perception, providing externally valid insights into how real-world factors influence people’s ability to detect and respond to false information in their daily lives.

Automated inspection of power infrastructure with drones requires processing video streams in real time and performing object recognition from image data with constrained resources. Server-based object recognition algorithms depend on transmitting data over a network and require considerable computational resources. In this study, we present an automated system designed to inspect power infrastructure using drones in real time. The proposed system is implemented on the Rockchip RK3588 platform and uses a lightweight YOLOv8 architecture incorporating a Slim-Neck model with a VanillaBlock module integrated into the backbone. To support real-time operation, we developed a digital video stream processing system (DVSPS) to coordinate multimedia processor (MPP)-based hardware video decoding, with inference performed on a multicore neural processing unit (NPU) using thread pooling. The system can navigate autonomously using a closed-loop machine vision system that computes the latitude and longitude of electrical towers to perform multilevel inspections. The proposed model attained an 84.2% mAP50 and 52.5% mAP50:95 with 3.7 GFLOPs and an average throughput of 111.3 FPS with 34% fewer parameters. These results demonstrate that the proposed method is an efficient and scalable solution for autonomous inspection across diverse operational conditions.

Mega highway projects are very sensitive to a number of geotechnical risks in the form of soil instability, seismic events, and environmental change, leading to costly failures. The current risk assessment methods are not capable of integrating the range of datasets—spatial data, temporal data, and real-time streams of data—and thus lack good levels of predictability, combined with a lack of response time. Thus, with a view to addressing these barriers, we introduce the new concept of Big Data Geotechnical Risk Assessment Model (BD-GRAM), which applies big data analytics and advanced machine learning algorithms in order to better predict and mitigate geotechnical risks for different scenarios. BD-GRAM combines various methods adapted to geotechnical data samples. The Graph Convolutional Networks (GCNs) are utilized for spatial and temporal data fusion, where complex spatial dependencies and temporal variations in soil properties, as well as samples of seismic data, are considered. GCNs, with the enhancement of attention mechanisms, have the ability to increase accuracy by as much as 20% compared with the conventional methods. A hybrid model by combining LSTMs and FEA, misleading synergistic use of physical laws, as well as temporal patterns of data pertaining to predictive accuracy, with an improvement of 25%. Near real-time processing on Apache Kafka &amp; Apache Spark enables near real-time continuous monitoring of risk with alerting on. SHAP (Shapley Additive Explanations) ensures interpretability of the model outputs, as well as transparency of the factors driving the risk predictions. Lastly, the system is scalable using GPU-accelerated TensorFlow to Run Masses of datasets &amp; samples. This fully integrated approach is optimized in this way to further enhance predictive accuracy and reduce false-positives and false-negatives, enhancing the speed and response of geotechnical risk assessment responses in real time. BD-GRAM represents an effective way to meet the early identification and mitigation of geotechnical challenges in a scalable data-driven manner for improved resilience and safety of large-scale infrastructures in any given highway scenario.

The proposed system, “Multi-Classification Detection on Live Video” is an intelligent computer vision–based platform designed to detect and classify multiple object categories in real-time video streams. The system processes live video input from cameras or video files and applies deep learning models to accurately identify and label objects across multiple predefined classes simultaneously. It supports real-time monitoring, automated detection, and visual annotation, enabling effective analysis of dynamic environments. The platform utilizes state-of-the-art convolutional neural network architectures such as YOLO, transfer learning–based models to achieve high-speed and accurate multi-class detection. Object classification and localization are performed frame by frame, ensuring consistent detection even under varying lighting and motion conditions. The system achieves an overall detection accuracy of up to 85-90%, with optimized inference speed suitable for real-time applications. A robust video processing pipeline handles frame extraction, pre-processing, object tracking, and result visualization. Detected objects are displayed with bounding boxes, class labels, and confidence scores. The system also supports recording and saving processed videos for further analysis. Performance evaluation is conducted using metrics such as precision, recall, F1-score, and FPS (frames per second). The application is developed using Python, OpenCV, and deep learning frameworks such as Yolov8, with a backend powered by Django for live streaming and control. The system is scalable and can be extended for applications including surveillance, traffic monitoring, smart cities, and security systems. Overall, the project demonstrates the effective use of deep learning and real-time video analytics for accurate multi-class object detection.

Recent advancements in Artificial Intelligence and eXtended Reality (XR) have laid solid foundations for the development of a new paradigm in industrial maintenance under the light of Industry 5.0 framework. This research presents the design, development, and implementation of an XR-enabled remote maintenance framework that integrates real-time video collaboration, AI-assisted guidance, and a persistent digital asset knowledge layer based on Asset Administration Shells for Maintenance and Repair Operations (MRO). By combining fine-tuned Large Language Models (LLMs) with immersive XR interfaces, the proposed framework enables technicians to interact with virtual representations of industrial assets, access contextual instructions, and receive expert support remotely in real-time. Through seamless integration of historical MRO data, digital twins, and real-time sensor streams, the system facilitates dynamic fault diagnostics and Remaining Useful Life (RUL) estimation. Therefore, the proposed approach is positioned as a Metaverse-aligned implementation, combining synchronous multi-user collaboration, digital–physical coupling through digital twins, and semantic interoperability. The framework is validated through two industrial case studies, demonstrating its feasibility and practical impact on maintenance efficiency and knowledge transfer. The findings position the Industrial Metaverse as a transformative enabler in the future of AI-driven machinery health monitoring.