Authors - Anna Berko-Boateng, Chalisa Veesommai Sillberg, Mika Saari, Pekka Abrahamsson Abstract - Electronic waste (e-waste) is one of the fastest-growing waste streams worldwide, and in many low-resource settings, informal recycling is performed under hazardous conditions with limited access to occupational safety information. Existing AI-based waste-recognition systems are typically designed for industrial or high-resource environments and do not adequately address the infrastructural, usability, and safety constraints of informal work contexts. To address this gap, this paper presents a lightweight Android application that uses multimodal artificial intelligence (AI) to support occupational safety among informal e-waste workers. The application enables users to capture images of e-waste components and receive structured safety guidance, including risk levels and handling instructions. Through the design, implementation, and field evaluation of the system, five meta-requirements were derived for AI-supported safety tools operating in low-resource environments. The system was evaluated through field testing at two informal recycling sites in Accra, Ghana, using representative low-cost smartphones. The findings indicate that the participants perceived the guidance as relevant and useful, while the interaction flow operated reliably across heterogeneous devices and user backgrounds. Beyond demonstrating feasibility, the study contributes transferable design knowledge on AI integration, device constraints, and safety-critical communication in low-resource socio-technical contexts.
Authors - Volodymyr Chumakov, Oksana Kharchenko, Zlatinka Kovacheva, Andrii Poberezhnyi Abstract - The Hilbert–Huang transform is considered. This method is compared to other known methods for handling nonstationary processes, specifically, the windowed Fourier transform and wavelet transform. The comparison is based on real data: the sound radiation of an Unmanned Aerial Vehicle using the example of a small Unmanned Aerial Vehicle, Phantom 4, and real electroencephalograms of a healthy and ill person. The advantages of using the Hilbert–Huang transform over the Hilbert transform are shown, because the latter is used for narrow-band processes. The possibilities of frequency extraction in the case of beats are noted. It is emphasized that Hilbert–Huang transform offers a more adaptive and data-driven approach, allowing it to reveal intrinsic components that traditional methods often obscure. In addition, this method provides a clearer physical interpretation of instantaneous frequencies, which is crucial for studying rapidly changing real-world signals.
Authors - Puwis Thiparapkul, Tuang Dheandhanoo, Panasuddhi Suddhiprakarn Abstract - Project management in game and animation production often faces challenges because generic tools do not align with their unique workflows. To enhance team collaboration, this study applies User Experience (UX) and Human-Computer Interaction (HCI) design principles to improve team workflows. The design is built upon real-world operations and an analysis of current free and paid industry tools, specifically aiming to optimize efficiency for beginners and small-to-medium-sized studios. We developed a domain-specific project tracking system on a demo site based on the newly synthesized "Waterfall Storm" concept, which balances structured administrative oversight with highly flexible, modular production phases. This conceptual architecture was derived directly from empirical user feedback and qualitative interview insights across three key target segments: entrepreneurs, creative practitioners, and students. The proposed design was evaluated through extensive empirical testing involving more than 100 users and 30 projects across educational and industrial environments. The results demonstrate that the Waterfall Storm workflow significantly enhances team collaboration, provides decentralized task visibility, increases clarity in asset tracking, and effectively eliminates format fragmentation while lowering software costs. These findings highlight that a domain-specific, user-centered approach to workflow design supports creative team collaboration more effectively than general-purpose project management framework.
Authors - Majdi Rawashdeh, Dhia Eddine Salhi, Awny Alnusair, Ali Karime Abstract - Ensuring student safety during school transportation remains a critical challenge, motivating automated, intelligent monitoring solutions. This paper introduces a comprehensive IoT-enabled framework for real-time student identication and attendance management aboard school buses. The proposed architecture combines an ESP32-CAM edge device with a suite of machine learning and deep learning models, evaluated on an augmented facial dataset of 40,000 images derived from the Labeled Faces in the Wild (LFW) benchmark. A comparative study of six pipelinesa basic SVM baseline, SVM with PCA, a distance-based classier, SVM with augmentation and grid search, Random Forest, and a proposed CNN-LSTM hybridis conducted. The CNN-LSTM hybrid achieves the highest accuracy of 99.5%, with precision, recall, and F1score exceeding 99%. The architecture spans four layerssensing, gateway, server, and applicationenabling low-latency communication between the edge device, cloud, and a mobile application serving parents and administrators, with end-to-end inference latency below 200 ms per frame. The results validate the system as a scalable, cost-eective, and highly accurate solution for modernizing student safety and attendance in school transportation.
Authors - Md. Istiaq Ahmed Bhuiyan, It Ee Lee, Teong Chee Chuah, Muhammad Sheraz, Gwo Chin Chung Abstract - Two-wheeled unsteady robots have special mobility benefits, but are unstable, nonlinear devices. Conventional control algorithms frequently fail to stabilize dynamic uncertainties and variations in the system. This study presents a Deep Reinforcement Learning (DRL) model based on Deep Q-Network (DQN) algorithm to balance autonomously. The proposed architecture was trained using DQN algorithm. It uses Exponential Moving Average filter to prevent high-frequency fluctuations and allows the motor output to be smooth. Simulation results demonstrate that the DQN controller successfully and robustly stabilizes the robot under mild to moderate initial pitch disturbances of up to 18°. However, boundary stress testing at an extreme 20° initial pitch revealed a critical kinematic limitation. The evaluation confirmed that while massive pitch recovery is algorithmically possible, the extreme actuator effort required to correct the chassis induces an uncontrollable divergence in the roll angle, leaving the system highly vulnerable to roll-axis instability.
Authors - Prince Kelvin Owusu, Moses Aggor, Gibson Afriyie Owusu, Emmanuel Mensah Azagadli, Martins Larweh Neurtey, Suleman Zakaria, Cecil Selorm Mensah Abstract - Campus Area Networks play a central role in supporting teaching, learning, administration, research, e-learning platforms, institutional databases, and communication services in higher education institutions. However, as universities become increasingly dependent on digital infrastructure, weaknesses such as poor network segmentation, inconsistent access control, insecure wireless access, limited monitoring, and service disruptions can expose institutional systems to unauthorized access and operational risks. This paper examines security and resilience challenges in the existing Campus Area Network at Pentecost University, Ghana, and presents a redesigned architecture based on VLAN segmentation, access-control enforcement, hierarchical network organization, and improved monitoring. The study adopts a mixed-method and technical assessment approach involving stakeholder input, technical observation, network audit, and pre/post evaluation of selected security and performance indicators. The redesigned architecture separates critical network zones, restricts unauthorized inter-VLAN communication, reduces unnecessary broadcast exposure, and strengthens the control of access to sensitive institutional resources. The results indicate that the intervention was associated with improved access control, reduced security incidents, lower latency and packet loss, increased throughput, and faster detection and mitigation of network incidents. The paper contributes practical evidence on how structured segmentation and access-control mechanisms can strengthen secure, resilient, and sustainable ICT infrastructure in higher education environments
