Computing and Computational Intelligence (CCI) 1
Track C6F1 CCI 6.1: Computing & Computational Intelligence (CCI) 6.1
Room: F1. Sipadan I (Level 4)
Chair: Rosalyn R Porle (Universiti Malaysia Sabah, Malaysia)
8:00 VoCare AI: a Multi-Agent LLM Workflow for Improved Clinic Operational Efficiency
Derrick Jo Han Lim, Pai Chet Ng and Malcolm Yoke Hean Low (Singapore Institute of Technology, Singapore)
In Singapore's polyclinics, touch-based self-service kiosks are widely used for administrative functions such as appointment scheduling and billing. However, these systems pose accessibility challenges for elderly patients, often resulting in increased staff workload and longer wait times. This paper presents \textit{VoCare AI}, a voice-first conversational assistant designed to streamline healthcare administrative tasks through multi-agent workflows powered by Large Language Models (LLMs). The system integrates a graph-based orchestration framework (LangGraph) with specialized agents that manage NRIC-based identity verification, appointment handling, billing queries, and general FAQ responses. It features a fine-tuned Automatic Speech Recognition (ASR) model adapted for Singapore-accented English (Singlish), trained on the IMDA National Speech Corpus to handle accent variability, code-switching, and elderly speech patterns. Evaluation results show that the best-performing ASR model achieved a Word Error Rate (WER) of 22.22%, while the Retrieval-Augmented Generation (RAG) module demonstrated strong performance with 1.00 groundedness and 0.92 retrieval relevance. By focusing on Singapore's linguistic diversity and incorporating localized authentication mechanisms, this work demonstrates how AI-driven conversational agents can enhance accessibility and operational efficiency in public healthcare environments.
8:15 QMADM-W: A Hybrid MADM Framework for Cloud Service Selection with Unavailable Data
P Navya (NIT Warangal, India); Sanjaya Kumar Panda (National Institute of Technology Warangal & NITW Techsammelan Private Limited, India); Rashmi Ranjan Rout (National Institute of Technology Warangal, India)
The rapid expansion of cloud computing has made it increasingly difficult for users to determine the most appropriate cloud service provider (CSP). The provider offers diverse services, typically assessed based on quality of service (QoS) attributes, including throughput, reliability, availability, latency, and response time. Researchers often present these QoS attributes in a decision matrix and apply multi-attribute decision-making (MADM) algorithms to evaluate and rank the CSPs. However, in practical scenarios, not all CSPs satisfy every QoS attribute, leading to unavailable performance measure values in a decision matrix. To address this challenge, we develop a hybrid MADM framework for CSP selection that handles an incomplete decision matrix. The framework integrates QoS-aware MADM (QMADM) algorithms, QTOPSIS-W and QVIKOR-W with attribute weights (QMADM-W). It employs three imputation techniques to determine unavailable performance values: minimum (min), maximum (max), and mean. The weights are derived using the analytic hierarchy process (AHP) and the analytic network process (ANP). Simulation results using the QoS for web services (QWS) dataset demonstrate the framework's effectiveness in QTPOSIS-W, with consistent and robust performance observed under the mean imputation technique through sensitivity analysis. The proposed algorithms offer a reliable solution for selecting an optimal CSP, even for an incomplete decision matrix.
8:30 From Encoded Features to Quantifying Disfluency: a Deep Learning Approach for Stuttering Severity Classification
Ashita Batra (Indian Institute of Technology, Guwahati, India); Manpreet Singh Saluja, Prarbdh Tiwari and Pradip Das (Indian Institute of Technology Guwahati, India)
Stuttering is a speech disorder marked by interruptions in the natural flow of speech, affecting millions globally. Accurately assessing its severity-categorized as mild, moderate, or severe is essential for guiding effective interventions and enhancing speech therapy outcomes. In this study, we address the task of stuttering severity classification using two publicly available datasets: FluencyBank Timestamped and LibriStutter. We extract diverse feature representations from Whisper, Wav2Vec2.0, and MFCCs to capture both learned and handcrafted speech characteristics. These features are evaluated using three deep neural network classifiers: CNN, ResNet50 and Transformer. Among the combinations tested, Whisper-Small embeddings paired with ResNet50 yielded the highest performance, achieving an overall accuracy of 82.1% and an F1 score of 0.778. While, the other Whisper variants also performs fairly competitive with an accuracy of 77.7% on Base and 76.6% on tiny model using ResNet50. Our proposed setup establishes a strong benchmark and a potential state-of-the-art-for future research in automatic stuttering severity assessment.
8:45 Person Re-Identification with Structural Semantic Graphs in Resource-Constrained Environments
Munir Bin Rudy Herman and Pai Chet Ng (Singapore Institute of Technology, Singapore); Olivia Shanhong Liu (Singapore University of Technology and Design, Singapore)
Person re-identification (ReID) aims to retrieve images of the same individual across disjoint camera views in surveillance networks, where resource constraints and privacy concerns often make traditional deep learning approaches impractical. This paper presents a lightweight, training-free ReID framework based on structural semantic graphs, designed for deployment in resource-constrained environments. Each person is represented by a multi-layer graph encoding body parts, fashion items, and dominant colors, extracted using pretrained pose estimation, fashion detection, and color clustering models. Without requiring any task-specific fine-tuning or GPU acceleration, our method enables interpretable matching through a hybrid similarity function that blends cosine similarity over weighted features with Jaccard similarity over attribute sets. Evaluated on the DukeMTMC-reID dataset, our method achieves 27.06% Rank-1 accuracy and 8.10% mAP, outperforming attribute-based baseline and random retrieval method. Further experiments validate that our hybrid similarity computation enhances retrieval performance and confirm that our bag-of-features graph encoding offers an effective solution for training-free ReID in resource-constrained environments.
9:00 Beach Litter Detection Using SwinTransformer Optimized Model for Drone Image
Poorna Pushkala K (Vellore Institute of Technology - Chennai Campus, India); Subbulakshmi P (Vellore Institute of Technology, India)
Beach litter detection (BLD) is the process of detecting and classifying abandoned materials in the beach environment. Drone image acquisition in BLD plays a crucial role in environmental conservation and waste management, particularly in unmanned coastal areas. Existing models primarily rely on traditional image classification techniques, which involve manually extracting features from the images. However, these methods often struggle with high false positives, low accuracy, and inefficiencies in processing large datasets from drone imagery. To overcome these limitations, this research proposes a new model, the Hybrid Swarm Gray Optimization-based Hybrid SwinNetCAM (SGO-SwinNetCAM) framework. The SGO-SwinNetCAM model contributes by integrating the Swin Transformer to capture both global and local features of the litter images, EfficientNetV2 for computational efficiency, and CBAM (Convolutional Block Attention Module) for refined feature representation. This hybrid combination optimizes performance in detecting and classifying beach litter. Additionally, the SGO algorithm fine-tunes hyperparameters, enhancing model accuracy and generalization. The proposed model aims to enhance the accuracy, efficiency, and robustness of BLD in drone imagery, even under challenging environmental conditions.
9:15 ActivityNet-HE: An Encryption-Enabled Deep Learning-Based Framework for Secure Human Activity Monitoring
Ahsanul Islam, Sadia Akter and Tahsina Farah Sanam (Bangladesh University of Engineering and Technology, Bangladesh)
An increasingly popular non-intrusive monitoring technique in smart environments is the recognition of human activities using Wi-Fi Channel State Information (CSI). However, the exposure of raw CSI data during processing poses significant privacy risks. This paper presents ActivityNet-HE, a deep learning framework designed to perform secure activity recognition through homomorphic encryption. The framework utilizes Principal Component Analysis (PCA) to minimize the dimensionality of the data makes it efficient to compute and extracts 29 statistical and signal-based features from time and frequency domains for each PCA stream. A lightweight neural network with polynomial activation enables encrypted inference without revealing sensitive input data. Evaluated on the CSI-HAR dataset, the model achieves 94.05% accuracy on both encrypted and plain data, demonstrating that strong privacy preservation can be achieved without sacrificing performance. While the system ensures secure inference, it incurs additional latency due to encryption overhead, which may limit its use in low-latency real-time applications. This makes the framework especially relevant for privacy-sensitive cases like monitoring the stereotypical motor movement (SMM) of autistic children in home or healthcare settings.
Control Systems & Robotics (CSR)
Track C6F2 ECD 6.1: Electronics, Circuits & Devices (ECD) 6.1
Room: F2. 501 Kadamaian (Level 5)
Chair: Bih Lii Chua (Universiti Malaysia Sabah, Malaysia)
8:00 Attentive Depth-Mapped Dice Loss for Accurate Segmentation of Smaller Objects in Medical Images
Dibakar Malakar, Joyita Bhattacharjee and Rinku Rabidas (Assam University Silchar, India)
Accurate segmentation of medical images plays an important role in clinical diagnosis, and has gained attention in recent years. Deep Learning based approaches are exploited for lesion segmentation in medical images due to its promising performance. In Deep learning, loss functions are crucial for proper training and evaluation. Dice loss, being simple and effective, is the most commonly used region based loss function for semantic segmentation, but has several limitations which includes insensitivity towards the boundaries and the distance between non-overlapping regions resulting in poor delineation of small lesions. To mitigate this issue, an Attentive Depth-Mapped Dice (ADD) loss function is proposed that incorporates distance maps derived from ground truth masks via Distance Transform and Signed Distance Transform, guiding the networks focus towards hard-to-segment boundary regions. Compared to Dice loss, the proposed approach achieves 24.32% and 12.44% improvement in HD95 scores, for small objects like Pancreatic Tumor and Hepatic Vessel, respectively. However, 10.04% increment is observed for Hepatic Vessel with minor improvement in pancreatic tumor in-terms of dice score.
8:15 Graphene-Based Bow-Tie Plasmonic Tweezers for Enhanced Optical Trapping at Terahertz Frequencies
Swapnil Siddiky (Bangladesh University of Engineering and Technology, Bangladesh); Ahmed Zubair (Bangladesh University of Engineering and Technology, Bangladesh & Rice University, USA)
Graphene-based plasmonic tweezers achieve strong localized field confinement by taking advantage of their low-loss properties and unique tunable conductivity. To improve near-field localization and optical trapping performance, we designed an optical nanotweezer structure in the shape of a bow-tie with a gap at the center. Using extensive parametric analysis, we explored structural dimensions, including gap size, length, and width, impact electric field, trapping performance, and localized surface plasmon resonance. Our findings demonstrated that while width primarily affected near-field strength, increasing length caused a redshift in the resonant frequency. We obtained a peak optical force of -1.57 nN/Wμm-2 at 7.3 THz frequency on a nanoparticle by introducing an 8 nm gap between the two bow-tie triangles, which is twofold higher than that of the gapless bow-tie structure. A deep optical potential well (> - 8x103 kBT/Wμm-2) was revealed close to the surface. The field and potential variation along the vertical axis further confirmed stable trapping. These findings highlight the potential of graphene plasmonic tweezers for high- precision dielectric and biological nanoparticle manipulation in lab-on-a-chip applications by demonstrating their strong geometrical tunability.
8:30 Preventive Maintenance Service (PMS) Scheduling: Development of Transformer Monitoring Device Considering Temperature and Vibration
GP Florianne Maghuyop Jabagat (Mindanao State University - Iligan Institute of Technology, Philippines); Marven E Jabian (MSU - Iligan Institute of Technology, Philippines & Mindanao State University - Iligan Institute of Technology, Philippines)
This study presents the development of a distribution transformer monitoring device purposefully designed to measure temperature and vibration, and its evaluation. The system constitutes a monitoring device energized by a solar-powered source scheme and a centralized data receiver. Its two-layered printed circuit board is customized through KiCAD, ensuring efficiency in component layout and optimized signal integrity. The enclosure design is made possible through Fusion 360, ensuring it features a secure mounting, appropriate placement of sensors, protection from the environment, and the inclusion of a solar charging system for autonomous operation. Accommodating the limitation of the NRF24L01 transceiver, the firmware implemented efficient data structures and its protocols for transmission. Sensing, storing, and organizing incoming data from multiple transmitters without collision is executed successfully by the centralized data receiver, aiding serial communication for the Human-Machine Interface (HMI). The garnered results proved the validity of the performance and consistency of the device subjected to changing circumstances. Consequently, the study infers that the monitoring device has the potential to serve as a solution for real-time condition monitoring of transformers, highlighting potential deployment in electrical distribution systems, thus contributing to enhancing maintenance practices and ensuring a reliable system.
8:45 Study of Insulative and Dissipative Material Surface Resistance Across Temperature and Humidity
Jeevan Kanesalingam, See Fung Lee and Hock Guan Ho (Motorola Solutions, Malaysia)
In the explosive atmosphere, and ignition can occur if a static spark comes into contact with explosive gasses or dust. An enclosure surface with a high surface resistance has a higher chance of surface electrostatic charge buildup due to the reduction of transferred charge. A reduction of transferred charge (increased surface resistance) would subsequently cause a higher electrostatic charge buildup on the surface; which increases the likelihood of a spark ignition in the explosive atmosphere. Surface resistance can vary with temperature and humidity (%RH). This paper studies the change of surface resistance of 3 different materials (PVC, ESD mat, PET) across temperature and humidity (%RH). Temperatures and humidity ranging from 15℃ - 70℃ and 30 - 90 % relative humidity (%RH) was used for the study. The paper confirms that % RH and surface resistance have an inverse relationship while the relationship between temperature and surface resistance is also having an inverse relationship.
9:00 Numerical Evaluation and Simulation-Based Performance Analysis of Transition Metal Dichalcogenide (MX₂; M=Mo, W; X=S, Se, Te) Heterojunction Photodetectors
Sourav Podder and Areeb Ahmed Zulkifle (Bangladesh University of Engineering and Technology, Bangladesh); Farseem Mohammedy (Bangladesh University of Engineering & Technology, (BUET)., Bangladesh)
A systematic numerical investigation is conducted on two-dimensional transition metal dichalcogenide (TMD)-based photodetectors comprising a heterostructure of p-CuO/MX₂/n-TiO₂, where MX₂ (M=Mo, W; X=S, Se, Te) denotes a family of layered semiconductors with tunable bandgaps. The device architecture incorporates a wide-bandgap window and back-surface layer to enhance carrier selectivity and suppress interfacial recombination. Key performance metrics, including photogenerated current density, spectral responsivity, and external quantum efficiency, are evaluated under standard solar illumination conditions. Parametric variations in absorber layer thickness and defect density are analyzed to assess their impact on photodetector performance. Among the studied materials, Tungsten Ditelluride (WTe₂) and Molybdenum Diselenide (MoSe₂) exhibit enhanced near-infrared sensitivity, achieving responsivity values up to 0.72 A/W. Furthermore, the spectral responsivity of the photodetectors was analyzed at various temperatures, where different active layer materials exhibited distinct trends. The results highlight the spectral adaptability and design tunability of TMD-based heterojunction photodetectors, positioning them as promising candidates for integration into broadband, high-performance optoelectronic platforms.
9:15 EEG Based Motor Imagery of Hands Classification Using Modified Sub-Band Common Spatial Pattern
Yash Madanlal Kashyap (Indian Institute of Technology Kharagpur, India); Eashita Chowdhury (Indian Institute of Technology, Kharagpur, India); C S Kumar (Indian Institute of Technology Kharagpur, India); Manjunatha Mahadevappa (Old NCC Building & Indian Institute of Technology Kharagpur, India)
Brain-computer interfaces (BCIs) leveraging motor imagery (MI) create intuitive systems that translate brain activity into direct control of computer applications. Motor imagery is a technique that provides simulated solutions during training for motor problems. For this study, Electroencephalography(EEG) data is collected from fourteen healthy subjects with 11 channels for two tasks with 40 repetitions each; the data were sourced from a publicly available database, ‘OpenVibe'. The present work proposes a modified Subband Common Spatial Pattern (mSBCSP) with Linear Discriminant Analysis as a method to classify right and left hand motor imagery movements, which provides a significant accuracy of 91.79% compared to other methods. The mSBCSP features are extracted from six EEG sub-bands after band pass filtering to enhance the discriminatory features. The results suggest that the proposed method offers a reliable and efficient framework for MI classification, which can serve as a foundation for future development of BCI-controlled applications for motor rehabilitation and the multimedia industry.
Power, Energy & Electrical Systems (PES)
Track C6F3 PES 6: Power, Energy & Electrical Systems (PES) 6
Room: F3. 502 Mesilau (Level 5)
Chair: Nur Ashida Salim (Universiti Teknologi MARA, Malaysia)
8:00 Galvanic Corrosion and Thermal Modelling of ACSR/TW Conductor in a Fully Immersed Acidic Industrial Electrolyte
Shahnurriman Abdul Rahman and Izzat Nawawi (Universiti Sains Islam Malaysia, Malaysia); Konstantinos Kopsidas (University of Manchester, United Kingdom (Great Britain)); Shamsul Fahmi Bin Mohd Nor (Universiti Sains Islam Malaysia, Malaysia & USIM, Malaysia); Norhidayu Rameli and Izzuddin Mat Lazim (Universiti Sains Islam Malaysia, Malaysia)
Galvanic corrosion is a critical degradation mechanism affecting the long-term performance of overhead conductors, particularly in industrial environments characterized by high acidic pollutants. This study investigates the coupled effects of galvanic corrosion and thermal behaviour in Aluminium Conductor Steel Reinforced (ACSR) with trapezoidal wires (TW) when fully immersed in an acidic industrial electrolyte. Using COMSOL Multiphysics, a comprehensive electrochemical-thermal simulation was developed to model the degradation and heat transfer characteristics of both pristine and corroded trapezoidal wire conductors, with the codename of Drake/TW. The analysis considers different pH levels (6.5 and 5.0) to reflect varying degrees of industrial acidity under a sustained operating temperature of 75 °C and 100 °C. At 100 °C operating temperature, simulation results reveal that, over a four-year period, aluminium loss of up to 62.4 mm² at pH 6.5 leads to a conductor temperature rise of 10.3 °C and a reduction in ampacity by approximately 85 A. Under more aggressive conditions at pH 5, corrosion accelerates, resulting in a greater temperature increase (16 °C) and ampacity reduction of up to 103 A. These findings highlight the substantial thermal impact of galvanic corrosion and emphasize the necessity of integrating corrosion evolution into ampacity assessments to ensure safe and reliable operation of overhead transmission lines in corrosive industrial environments.
8:15 Comprehensive Analysis of Solar Power Generation: Time-Series Dynamics and Clustering Patterns for Predictive Modeling
Hasanur Zaman Anonto (American International University-Bangladesh, Bangladesh); Md Ismail Hossain (American International University Bangladesh, Bangladesh); Afrin Tanzila Rabbani, Samiul Ahasan Sajid, Abu Shufian and Toriqul Islam (American International University-Bangladesh, Bangladesh); Md Sultan Mahamud (Lamar University, USA)
The growth of the renewable energy target has contributed to increasing the needs of a powerful solar generation system. But those caused by solar power production itself from irradiation, temperature, and the state of the solar panels present tough challenges in terms of prediction, beneficial use and optimization. The objective of this work is to model the solar power generation forecasting accurately using machine learning models like Linear Regression, Decision Tree Regressor and Random Forest Regressor. The work investigates the influence of the environment on the power output, underlining the diurnal variability of irradiation and temperature daily. Here, it describes how the use of time-series modelling and clustering algorithms help identify patterns of solar power behavior distilled from data collected across multiple months and locations. The results show that compared to other models the Random Forest model achieved a boosted accuracy rate of forecasting by capturing non-linear overheads and minimizing the variance. The best-case scenario yields 92.5W peak DC power generation and the worst-case scenario shows 12W minimum peak DC power generation, and solar power generation is heavily affected by temperature and irradiation changes as in daily temperature fluctuations from 20°C to 85°C in addition to combining anomaly detection using Z-Score analysis and Isolation Forest algorithms to identify problems such as degradation of panels and fault of sensors. These findings emphasize the necessity for a near real-time model and complex forecasting models to mitigate issues associated with renewable energy precursors to increase solar energy productivity and efficiency.
8:30 Technical and Economic Assessment of Renewable Energy Integration into Saudi Arabia's Fossil Fuel Based Grid
Zaid S. Al Otaibi (King Abdulaziz City for Science and Technology (KACST), Saudi Arabia); Mazen A. Baabbad and Saad S. Alshahrani (King Abdulaziz City for Science And Technology, Saudi Arabia)
Saudi Arabia faces a pivotal moment in its energy transition, aiming to diversify electricity generation while remaining a top energy exporter. With abundant solar and wind resources, the Kingdom has strong renewable potential but must overcome operational, economic, and infrastructure challenges. Electricity transmission, especially over long distances, involves losses and high costs, with trade-offs between HVAC and HVDC technologies. This study examines these costs alongside utility-scale renewable projects, highlighting solar PV and onshore wind's growing cost-competitiveness but also their intermittency and low capacity factors that limit grid efficiency. Fossil-based dispatchable generation, like Open Cycle and Combined Cycle Gas Turbines, remains crucial for reliability. Using California as a case study, the paper reveals barriers such as curtailment, seasonal variability, and limited storage that challenge full renewable reliance. Given that fossil fuels still generate 80% of Saudi electricity, transitioning to a hybrid system demands region-specific modeling of generation, transmission, and storage. This research offers a framework to support Saudi Arabia's energy diversification and carbon reduction goals sustainably.
8:45 Data-Driven Optimization of Wind-Solar Tower Performance a Machine Learning Approach to Thermal Updraft Prediction
MD. Minhajul Islam (American International University - Bangladesh, Bangladesh); Ahmed Intekhab Rohan (Islamic University of Technology, Gazipur, Bangladesh); Hasanur Zaman Anonto (American International University-Bangladesh, Bangladesh); Md Ismail Hossain (American International University Bangladesh, Bangladesh); Md Mukter Hossain Emon and Md Akteruzzaman (Lamar University, USA); Abu Shufian (American International University-Bangladesh, Bangladesh); Md Siddikur Rahman (Universiti Teknologi PETRONAS, Malaysia)
This paper aims at examining how machine-learning can be applied to achieve optimal wind-solar hybrid towers performance in relation to predicting thermal updraft. Wind-solar towers use solar irradiance and wind speed to generate energy but due to the complex nature of interaction between the various environment variables, wind-solar towers are relatively hard to predict the thermal updraft due to the interaction between the solar irradiance, air temperature, and wind velocity. A complete dataset that covers 90 days is used, and various machine-learning models, such as linear regression, support vector machines, and neural networks, are tested to make a prediction of thermal updraft. The findings show that among all the alternative models, linear regression had a much superior predictability and efficiency in computation. The obtained result showed a close-to-perfect fit, R2 being equal to 0.98 and indicating the extremely high ability of the regression model to make predictions of the updraft velocity under heterogeneous environmental conditions. Sensitivity analysis had demonstrated that solar irradiance and wind speed are the best predictors of thermal updraft thereby offering insights that would be adopted to streamline the hybrid renewable energy system, and energy forecasting with potential to boost power generation especially during days with low solar irradiance.
9:00 Data-Driven Identification of VHIF-Prone Areas Using Waveform Behavior and Unsupervised Learning
Sophia Mae M Gascon (Mindanao State University-Iligan Institute of Technology, Philippines); Rovick Tarife (Mindanao State University- Iligan Institute of Technology, Philippines)
This paper presents a data-driven approach for identifying high-impedance fault (HIF)-prone areas in power distribution networks. The study focuses on staged vegetation-related HIFs (VHIFs) drawn from the publicly available VeHIF dataset. Recognizing the scarcity of real-world HIF data and the absence of precise location labels, the method simulates fault occurrence across the IEEE 14-bus test system by probabilistically assigning VHIF cases to distribution network. High-frequency voltage signals are analyzed to extract discriminative waveform features, including wavelet-based coefficients and power spectral density (PSD) peaks, which are aggregated per line to capture behavioral risk patterns. K-means clustering is then applied to group lines into low-, medium-, and high-risk categories, enabling unsupervised classification of VHIF-prone regions. The results demonstrate that even without location-specific training labels, meaningful distinctions in VHIF risk levels can be inferred from signal behavior. This framework provides a scalable and interpretable diagnostic tool to support proactive fault management and wildfire mitigation strategies in data-limited environments.
Electronics, Circuits & Devices (ECD)
Track C6F4 ECD 6.2: Electronics, Circuits & Devices (ECD) 6.2
Room: F4. 503 Dinawan (Level 5)
Chair: Maizatul Zolkapli (Universiti Teknologi MARA, Malaysia)
8:00 FPGA Implementation of Lorenz System Based RO-PUF
Kota Naga Srinivasarao Batta (NIT Warrangal, India); Hemanth Reddy Kattamanchi and Mahesh Vanga (National Institute of Technology Warangal, India); Suresh Babu Erukala (Assistant Professor, India)
This paper presents the design and implementation of a Lorenz chaotic system-based Ring Oscillator Physically Unclonable Function (RO PUF) on Field Programmable Gate Arrays (FPGAs). PUFs have evolved as an important component of hardware security, providing unique IDs derived from inherent process variations during integrated circuit fabrication. The chaotic dynamics of the Lorenz system, which is recognised for its unpredictability and sensitivity to initial conditions, give a unique technique to creating random and secure challenges for PUF. The present research demonstrates the viability of incorporating chaotic systems into PUF designs to enhance security. The proposed system offers a reliable and adaptable way to generate secure and distinctive IDs that can be applied to a range of embedded systems, IoT, and cybersecurity applications. The results emphasise the potential of chaotic dynamics to considerably enhance the security features of hardware-based cryptographic primitives. By utilising the intrinsic variability of the Lorenz system, the proposed design optimises the security features of the PUF, guaranteeing a high degree of unpredictability and uniqueness. The experimental results indicate that the design is relatively low in hardware overhead, has a reliability of 97.79%, and a uniqueness of 48.91%. Additionally, the design parameters can be adjusted to alter the output response by the design parameters.
8:15 Biasing Schemes & Loading Effects in Analog Selectors Built with Transmission Gate Logic
Md Al Amin Ashik, Toufiq Mohammad Hasan, Jahidul Islam, Tamima Tabassum and Nafisa Noor (North South University, Bangladesh)
Transmission gate logic (TGL) is a popular circuit connection technique for analog selectors. Despite the numerous IC models and existing literature on simulation-based performance analysis for TGL-based selectors, the biasing strategies and loading effect problems are still not well discussed or characterized. In this work, we present these two vital issues with detailed SPICE circuit simulations and measurements conducted on hardware circuits. We have observed that 4-terminal MOS transistors with fixed body bias and face-to-face arrangement of parallelly connected NMOS and PMOS transistors are the only viable biasing schemes for TGL-based selectors. We have monitored considerable loading effects for load resistances lower than approximately 10 kilo-ohms from simulations and characterization of 1:2 and 1:4 demultiplexers and 2:1 and 4:1 multiplexers. We have discussed the underlying reasons behind the loading effect by analyzing the overdrive voltages, operation regions, drain-to-source resistances, and body effects of the NMOS and PMOS transistors at the active channels of selectors.
8:30 Design of a Compact Octagonal Patch Antenna with Defected Ground for Multiband 5G Applications
Md. Omar Faruq and Afzal Hossain (Aviation and Aerospace University Bangladesh, Bangladesh); Pran Kanai Saha (Bangladesh University of Engineering and Technology, Dhaka, Bangladesh)
This paper presents a compact octagonal patch antenna featuring two slots on the top and a central slot, designed to achieve multiband operation within the 5G frequency spectrum. A circular defected ground structure is incorporated, and the antenna is fed using a microstrip line. The design and analysis are conducted using CST simulation software, examining the effects of varying the sizes of three circular slots on the patch along with the defected ground. The proposed antenna operates at eight resonant frequencies within the 0-30 GHz range, specifically at 3.3, 3.8, 5.28, 7.8, 13.92, 16.56, 19.89, and 24.81 GHz. Key performance characteristics such as S-parameters, impedance response, and both 2D and 3D radiation patterns are analyzed. The S-parameter obtained from the simulation is experimentally verified, showing good agreement with the simulation results. With a compact size of 40mm × 40mm, this antenna is well-suited for FR1, FR2, and future 5G applications.
8:45 Inductive/Capacitive Coupling Hybrid WPT for Implantable Medical Device: Basic Analysis of Coupler Structure
Miyu Kodama and Dairoku Muramatsu (The University of Electro-Communications, Japan)
Wireless power transfer (WPT) is rapidly expanding beyond consumer and industrial applications and is now being investigated for wirelessly powering implantable medical devices (IMDs). Efficiency losses caused by coil misalignment in inductive links and electrode spacing in capacitive links motivate a millimetre-scale inductive/capacitive hybrid coupler (ICCH-WPT). In this study, we design and evaluate an ICCH-WPT coupler designed for abdominal IMDs such as abdominal neurostimulators. Electromagnetic field simulations using a multilayered human abdominal model show that a 7.5 mm × 7 mm hybrid coupler can deliver 5.85 mW at 1 MHz-66% higher than a capacitive-only design-while complying with human safety guidelines on electromagnetic exposure. The field energy is strongly confined near the coupler, minimizing deep-tissue exposure and indicating excellent biological safety. ICCH-WPT therefore satisfies the 3-5 mW power budget of modern neurostimulators. This makes it a promising solution for next-generation, wirelessly powered implantable devices requiring stable and safe wireless energy links.
9:00 Condition Monitoring for Induction Motors: Vibration and Temperature Data Acquisition for Predictive Maintenance Analysis
Godfrey F Mulit (Mindanao State University - Iligan Institute of Technology, Philippines & College of Engineering, Philippines); Harreez Villaruz (Mindanao State University - Iligan Institute of Technology, Philippines); Marven E Jabian (MSU - Iligan Institute of Technology, Philippines & Mindanao State University - Iligan Institute of Technology, Philippines)
Induction Motors are the backbone and workhorse machines in various industry sectors. Primarily used to provide increased efficiency and productivity in many industrial processes, minimizing equipment downtime and life-cycle improvement are of utmost importance. This study focuses on the design and development of a non-intrusive, microcontroller-based condition monitoring device for the vibration and temperature operating parameters of induction motors.
In this study, hardware and software systems are developed. An accelerometer module and infrared temperature sensors are integrated with ESP32-based development boards, as well as the application of radio frequency transceiver modules for wireless data transmission. Printed circuit boards are designed and fabricated alongside a 3D-printed enclosure to house all of the hardware components. Data acquisition techniques for the vibration and temperature parameters are implemented and transmitted through a combination of ESP-NOW and radio frequency communication protocols. These features allow real-time condition monitoring of induction motors even without physical intervention on the internal components of the machine. The condition monitoring device is tested on a developed vibration generator platform under different operating conditions with the use of a digital PWM control. The prototype device is shown to achieve all the intended functions, as well as provide potential for scalability. This paper recommends future studies to incorporate optimization on power management, integration of different communication protocols, reduction of device footprint, as well as field trials for outdoor, weather-exposed applications.
9:15 A Cost-Effective Approach to PCB Structure and Process for Enhanced High-Speed Signal Integrity
Jeonghyeon Choi and Youbean Kim (Myongji University, Korea (South))
As the use of high-speed signals continues to grow, precisely controlling signal loss has become an increasingly challenging task in the electronics industry. This paper focuses on dielectric loss that occurs in transmission channels during high-speed signaling and proposes a novel printed circuit board (PCB) structure to effectively mitigate such loss. To evaluate the feasibility of the proposed structure, electromagnetic simulations were conducted to analyze the electrical characteristics of the PCB. Furthermore, in order to reduce additional manufacturing costs, a practical PCB fabrication method that does not require high-end equipment was developed, and a prototype was manufactured. Finally, using a vector network analyzer (VNA), the electrical performance and signal quality improvement of the prototype were thoroughly evaluated. The experimental results demonstrate that the proposed structure can effectively reduce dielectric loss in the transmission channel while minimizing the need for expensive low-loss dielectric materials, making it a cost-efficient solution for high-speed PCB applications.
Communication Systems (CS) / Control Systems & Robotics (CSR
Track C6F5 CS 6 / CSR 5: Communication Systems (CS) 6 / Control Systems & Robotics (CSR) 5
Room: F5. 504 Madai (Level 5)
Chair: Azwan Mahmud (Multimedia University & Telekom Malaysia, Malaysia)
8:00 Distributed Algorithms for Age Optimal Scheduling in Multi-Hop Wireless Networks
Ramakrishnan K (Indian Institute of Technology Tirupati, India); Teena Mary Treesa (IIITDM Kancheepuram, India); Premkumar Karumbu (Indian Institute of Information Technology Design and Manufacturing Kancheepuram, India)
In this work, we consider a scheduling problem for a multihop network with multiple flows. For each flow, the source nodes generate packets that carry time-sensitive information to the destination, and the goal is to achieve a scheduler that has the minimum expected weighted sum age of information. This problem has significance in real-time systems such as IoT, industrial process control, and remote monitoring, where timely updates are more important than throughput or latency. Centralized age optimal scheduling policies are often impractical for large-scale networks due to the overhead and coordination required. We propose a distributed adaptive gossip-based scheduling algorithm for achieving minimum age. We compare our proposed scheme with state of the art distributed algorithms such as Fresh-CSMA and Greedy Maximal Scheduling, highlighting the trade-offs between performance and overhead. Through extensive simulations, we demonstrate that our algorithm achieves better performance under a wide range of network conditions and traffic loads while maintaining lightweight communication complexity.
8:15 Biometric Parameter Selection Towards Human Identity Recognition Using Wi-Fi CSI Sensing
Antoinette R Anastacio, Joshua Jake M Benitez, Ryan Carlo DJ Enriquez, Miguel Albert P Icuspit, Ken Wesley O Lampa, Azel Monique O Pangan and Josyl Mariela Rocamora Reyes (University of Santo Tomas, Philippines)
Traditional human activity recognition (HAR) systems utilize vision and wearable sensors to sense and identify human movements. When applied for security and surveillance, these HAR systems can also be utilized for human identity recognition (HIR), where the sensing technologies authenticate users. Recently, Wi-Fi sensing using channel state information (CSI) has emerged as a promising alternative technology for HIR applications, wherein the users are typically identified by their gaits as the primary biometric parameter. As there are other possible biometric parameters, this study explores Wi-Fi CSI sensing in HIR where the subjects perform three physical (static) and two behavioral (dynamic) biometric parameters. A CSI dataset involving samples collected from six subjects in a controlled indoor environment was preprocessed using discrete wavelet transform (DWT) and principal component analysis (PCA). Different parameter-specific HIR models were trained and tested and the HIR model based on the ‘draw triangle' behavioral parameter achieved the highest HIR accuracy of 98.61%. Results show that the five investigated gestures are viable biometric parameters alternative to gait.
8:30 Implementation of EGWO and ISCO Algorithms for Minimizing Transmission Power in IoT Clusters with Full Connectivity
Kotte Sowjanya (Kakatiya Institute of Technology and Science, India); Satish Kumar Injeti (National Institute of Technology Warangal, India & NITW, India)
In this day and age of the Internet of Things (IoT), it is of the utmost importance to guarantee a connection that is both stable and efficient in terms of energy consumption among sensor nodes. Wireless technology researchers and developers
I frequently come into a situation in which the energy requirements of sensors and the longevity of their total communication are in conflict with one another. As a consequence of this, a number of researchers are focusing their attention on locating the best possible answer. Improved Grey Wolf Optimization (IGWO) and Improved Sine Cosine Optimization (ISCO) are two algorithms that are developed, implemented, and utilized in this work. The purpose of these algorithms is to reduce the amount of energy that is consumed by a sensor node while simultaneously ensuring that all nodes are completely interconnected. By applying the IGWO and ISCO algorithms, the findings that were obtained suggest that the proposed technique produces superior results in comparison to those that were discovered in the literature. This is particularly true in terms of the amount of energy that is saved and the dependability of connectivity. EGWO and ISCO-based solutions were presented as a means of reducing the amount of power that is consumed by wireless sensor Internet of Things networks, according to the findings of the article.
8:45 Integrated Circuit Binning Count System for Semiconductor Manufacturing
Mark Angelo T Mercado, Oliver A. Medina and Juliet O. Niega (National University, Philippines)
The semiconductor industry in the Philippines accounts for nearly 60% of the country's total exports, playing a critical role in the national economy. However, many production facilities continue to depend on outdated system particularly those operating on legacy Windows NT platforms with decentralized, manual processes for Integrated Circuit (IC) binning. This study presents the design and implementation of an Integrated Circuit Binning Count System (ICBCS) that leverages microcontroller-based hardware to automate and centralize binning data acquisition. The system interfaces with legacy testing equipment to monitor IC classification pulses in real time, display local counts, and wirelessly transmit data to a centralized web-based server for storage, analysis, and reporting. The proposed solution significantly improves traceability, reduces human error, and enhances operational efficiency. Designed as a low-cost, scalable system, ICBCS supports the semiconductor industry's transition toward Industry 4.0 standards. Field testing in an actual manufacturing environment demonstrated its effectiveness, ease of integration, and adaptability to various production line configurations.
9:00 Constrained GA Optimized Super Twisting Sliding Mode Controller for 2-DoF Robotic Arm
Vishal Mehra (Gujarat Technological University & Anand Agricultural University, India); Dipesh Shah (Gujarat Technological University, India); Axaykumar Mehta (Institute of Infrastructure Technology Research and Management, India); Siddharth Mehta (Arizona State University, Tempe, AZ, USA)
This paper presents a constrained optimization method for the Super-Twisting Algorithm using a Genetic Algorithm. The Super-Twisting Algorithm is an important class of higher-order sliding mode strategy commonly employed in designing robust controllers and observers that effectively mitigate the chattering effect observed in conventional sliding mode control. The performance and stability of the dynamical system critically rely on the appropriate selection of gain parameters, so it is necessary to choose the optimal gains for the controller. A novel approach to optimize the control gain parameters using a constrained Genetic Algorithm is proposed to achieve the desired dynamic response. The constraints are derived from a strict Lyapunov function, thereby ensuring the finite-time stability of the closed-loop system through Lyapunov stability criteria. The performance of the proposed optimized gains is evaluated through real-time experiments on a 2-degree-of-freedom serial flexible joint robotic arm. The experiment results show that the proposed technique performs better regarding dynamic response, disturbance rejection, and chattering reduction.
9:15 Data Driven Controller for 6-DOF Quadcopter
Harikrushn Surani (M.Tech, India); Nikita Joshi (IITRAM, Ahmedabad, India); Axaykumar Mehta (Institute of Infrastructure Technology Research and Management, India)
This paper introduces a data-driven Artificial Neural Network (ANN) controller for a 6-DOF quadcopter, leveraging advanced machine learning techniques to address limitations in traditional control systems. The proposed data-driven controller eliminates the need for a mathematical model of the system, relying solely on accurate data for its design and functionality. The controller is trained on real-time data captured via OptiTrack motion capture and onboard sensors, ensuring precise position, orientation, and velocity tracking. Experimental results demonstrate the controller's superior performance in both hovering and trajectory tracking tasks, achieving greater stability, adaptability, and efficiency compared to the PIV controller. The data-driven approach exhibits faster convergence and enhanced robustness in maintaining position and following predefined trajectories under dynamic conditions. Additionally, the system demonstrates strong resilience to environmental disturbances. These findings underscore the potential of neural network-based controllers as a transformative solution for UAV operations, marking a significant advancement in the autonomy, reliability, and intelligence of modern aerial systems.
9:30 A Trust-Based Blockchain Framework for Mitigating Smartphone Theft Incidents
Suresh Babu Erukala (Assistant Professor, India); Banoth Krishna Mohan Naik (National Institution of Technology, Warangal, India); Aswani Devi Aguru (SRM University AP, India); Erukala Sudarshan (Sumathi Reddy Institute of Technology for Women (SRITW), India)
In today's digital age, mobile phones are vital for communication, navigation, and productivity, but their growing role as data repositories makes them prime targets for theft, leading to financial loss, privacy breaches, and operational disruptions. Traditional prevention methods often lack robust data protection and are vulnerable to unauthorized access. To address these challenges, this paper proposes a blockchain based mobile theft detection system using Hyper ledger Fabric, a permissioned blockchain platform. The system assigns each mobile device a unique identifier (UID) upon registration and securely stores metadata, such as location and status, on the blockchain. In case of theft, the owner can report the incident, triggering smart contracts to update the device's status and notify network participants in real time. This decentralized approach ensures that stolen devices are flagged and blocked across participating networks. Additionally, the system supports secure ownership transfer, reducing fraud and enhancing trust in the mobile ecosystem.
Computing & Computational Intelligence (CCI) 2
Track C6F6 CCI 6.2: Computing & Computational Intelligence (CCI) 6.2
Room: F6. 505 Sepilok (Level 5)
Chair: Khairil Anuar (Multimedia Universiti, Malaysia)
8:00 EDSGAN: an Edge-Informed Generative Adversarial Network for Enhanced Perceptual Quality Single Image Super-Resolution
Masuma Aktar (National Institute of Technology, Silchar, India); Kuldeep Singh Yadav (NIT Silchar, Assam, India); Rabul Hussain Laskar (NIT SILCHAR, India)
Single Image Super-Resolution (SISR) faces a persistent challenge in reconstructing high-frequency edge details, which are paramount for human perceptual quality. While Generative Adversarial Networks (GANs) have significantly advanced SISR, they often struggle to generate truly sharp and realistic edges, often due to their inherent loss functions. To address this critical limitation, we propose an Edge-Informed Super-Resolution GAN (EDSGAN). EDSGAN employs a dual-path edge-informed discriminator that simultaneously analyzes image content and edge maps, enabling more effective discrimination of realistic versus artifact-prone edge structures. Concurrently, an innovative edge-aware loss guides the generator towards reconstructing perceptually sharper and more accurate edges. Extensive experiments on benchmark datasets demonstrate EDSGAN's superior perceptual quality. Notably, on the Set14 dataset, EDSGAN achieves a remarkable improvement of 9% in LPIPS (from 0.1329 to 0.121) and 8% in PI (from 2.9261 to 2.689) over the ESRGAN baseline. Our method effectively strikes an optimal balance between visual realism and pixel-level accuracy.
8:15 Deep Learning Based Tree Counting Method for Various Plant Density in UAV-Captured Palm Oil Plantation Images
Eisyatul Hannie Binti Mohammad Rushdan (University Putra Malaysia, Malaysia); Syamsiah Mashohor and Khairunniza Bejo (Universiti Putra Malaysia, Malaysia)
Oil palm is a globally significant economic crop, especially in Malaysia and Indonesia, where it contributes substantially to national economies. Accurate monitoring of plantations, particularly tree count and health, is essential for effective management and yield estimation. However, traditional field-based methods are costly and labor-intensive, and while UAVs and very high-resolution satellite imagery offer precision, they are often limited by high cost, limited coverage, and technical constraints such as altitude variation and image mosaicking. This study proposes a scalable and cost-effective pipeline that utilizes UAV imagery implementing two tree-counting approaches Template Matching, serving as a classical baseline, and YOLOv8, a deep learning object detection model chosen for its high accuracy and inference speed. The oil palm tree density was classified into High, Medium, and Low category for easy result observation. Preliminary results demonstrate promising True Positive Rate (TPR), False Negative Rate (FNR), and Estimation Error (EE) values, with YOLOv8 achieving its best performance in medium-density regions (TPR: 109.54%, FNR: 9.50%, EE: 12.91%) and Template Matching showing the weakest performance in low-density areas (TPR: 305.45%, FNR: 205.45%, EE: 205.45%), indicating that the proposed approach offers a practical, real-time, and resource-efficient solution for large-scale oil palm monitoring, supporting sustainable plantation management.
8:30 TigerNav: Development of a Virtual Assistant Using an Autoregressive Model for Indoor Navigation
Ralph Alexander N. San Juan, Ernest John Q. Baetiong, Saranggani J. Bantayao Jr, Marc Justin M Mangali, Carl Kristien P Sumo and Ma. Madecheen S Pangaliman (University of Santo Tomas, Philippines)
Indoor navigation systems play a crucial role in guiding users through complex environments such as airports, shopping malls, hospitals, and university campuses. With the advancement of Artificial Intelligence (AI), researchers have explored dialogue-based approaches to enhance user interaction within these systems. However, existing models face limitations in accurately processing diverse user inputs, which poses challenges in efficiency and usability. To address this, the researchers developed a virtual assistant using a Large Language Model (LLM) to facilitate a dialogue-driven navigation experience. The study was conducted in three phases: (1) generation of datasets, (2) development of models, and (3) virtual assistant performance assessment through a user satisfaction survey. The dataset, consisting of 200,000 entries, was split into 70% for training and 30% for testing, designed to train the navigation model on real-world scenarios. Due to hardware limitations, specifically the 4GB of VRAM on NVIDIA GeForce GTX 1650 and RTX 3050 GPUs, the GPT-2 model was selected as the base model. Despite being an outdated and less capable model for handling complex language tasks, GPT-2 was chosen due to its compatibility with limited hardware resources, such as GPUs with only 4GB of VRAM. Among the training methods tested, the General Purpose Trainer achieved the best performance, with a BERT score of 0.89, a METEOR score of 0.84, and an average perplexity of 2.82, indicating moderate success in understanding and responding to user input. A user satisfaction survey further validated the practicality of the system, with an average rating of 87.44%. Although the model shows promise, its effectiveness in interpreting highly unstructured or unconventional queries could be improved using more advanced LLMs.
8:45 Pestaway: a Pest Management Chatbot with Integrated Speech Capability
Rachel Hannah C. Dela Cruz, Joshua Carlo C Aguarin, Nickolas Chase P Ling, Maveric S. Magsaysay, Jastin Brylle C Villanueva and Ma. Madecheen S Pangaliman (University of Santo Tomas, Philippines)
Farmers faced challenges with pest control, leading to lower crop yields and financial losses due to limited access to accurate information. To address this, the study introduced Pestaway, an AI-powered chatbot for Filipino farmers that uses Large Language Models (LLMs) to provide interactive and speech-enabled solutions. A data set of 80,000 pairs of questions and answers, equally divided between English and Tagalog, was generated using GPT-4 and credible agricultural sources. The models were fine-tuned using different training approaches: Supervised Fine-Tuning (SFT), SFT with Direct Preference Optimization (DPO), and Odds Ratio Preference Optimization (ORPO). Among the evaluated trainers, the ORPO Trainer produced the best results, achieving a METEOR score of 0.3271, a BERTScore of 0.7316, and an average perplexity of 1.5185. Additionally, the metrics word error rate (WER) and character error rate (CER) were used to evaluate the speech recognition capabilities of the chatbot. The WER ranged from 0% to 28.57%, which showed that some transcriptions exceeded the acceptable threshold, while the CER ranged from 0% to 3.43%, which remained low across tests, indicating more accurate transcriptions due to minimal character-level errors. These findings demonstrated that the chatbot was able to recognize voice queries in English and Tagalog with reliable precision. Furthermore, a user satisfaction survey involving farmers, agricultural students, and individuals interested in agricultural practices yielded an average rating of more than 4.1 out of 5, with 97.73% giving positive feedback and 100% recommending the chatbot for use. The participants highlighted the usefulness, precision, and bilingual capability of the chatbot and suggested additional enhancements such as faster responses, visual aids, and a mobile version.
9:00 Enhancing Software Refactoring Prediction Accuracy Through Feature Selection and Data Sampling Strategies
Aditya Kumar Singh Bisht (NIT, Kurukshetra, India); Lov Kumar and Vikram Singh (National Institute of Technology, Kurukshetra, India)
Accurate classification and prediction, especially in critical software engineering tasks like software refactoring, often face significant challenges due to the messy nature of real-world data: high dimensionality, irrelevant noise, and wildly imbalanced categories. This study introduces a robust and practical framework designed to mitigate these issues. We thoroughly examine how advanced feature selection methodologies, including statistical tests, correlation filters, and Principal Component Analysis (PCA), synergize with the Synthetic Minority Over-sampling Technique (SMOTE) for data balancing. Our approach involves systematically applying these preprocessing steps to 30 distinct real-world datasets, followed by evaluation with different and diverse machine learning models. Our findings are clear: SMOTE consistently and significantly boosts the reliability and performance of our models, particularly in identifying crucial minority cases, which is vital for effective refactoring prediction. Moreover, intelligent feature selection proves instrumental in optimizing performance, frequently leading to simpler models without compromising accuracy. Ultimately, Random Forest, Support Vector Machines, and Neural Networks emerged as the consistent top-performing machine learning models. This research offers practical, empirically proven strategies for building highly accurate, efficient, and stable predictive models, essential for modern software engineering and refactoring efforts.
9:15 Illegitimate Data Flow Detection of Data-Driven Applications Using Dependency Graph
Anwesha Kashyap (Indian Institute of Information Technology, Guwahati, India); Angshuman Jana (IIIT Guwahati, India)
Detecting potential data breaches in software systems can be effectively achieved through language-based information flow security analysis. Significant risk to confidential data, potentially leading to security breaches, is posed because of poor coding practices in database applications. To tackle the issue, our paper presents a dependency graph-based approach that identifies potential confidentiality leaks in software code. For a variety of software engineering tasks, including information flow security analysis, debugging, code optimization, enabling code reuse, and enhancing program comprehension, Dependency information plays an essential role. Existing dependency information techniques fail to generate precise results. Through the utilisation of syntax-based dependency analysis, we enhance and improvise on existing techniques by introducing Indirect DD-Dependencies, which can result in indirect information leakage. A precise dependency information, which improves the accuracy of security analysis results, is provided through this refinement. We provide experimental results on benchmark codes to validate our approach. To determine illegal information flows, we use propositional formulae for the security-level truth assignments. Thorough utilisation of such an approach aids in bridging the gap between existing approaches and future applications to handle the same.
Engineering Technologies & Society (ETS)
Track C6F7 ETS 6: Engineering Technologies & Society (ETS) 6
Room: F7. 506 Selingan (Level 5)
Chair: Mohd Suffian bin Misaran Status (Universiti Malaysia Sabah, Malaysia)
8:00 Ethical Design Framework for Enhancing Accessibility in Graphic Platforms for Colorblind Users
Michael P Camacho and John Heland Jasper Ortega (FEU Institute of Technology, Philippines)
Color is a foundational element in digital design, yet individuals with color vision deficiency (CVD) often encounter significant barriers when engaging with mainstream graphic design platforms. This study explores the ethical and practical implications of designing for users with CVD, focusing on the experiences of Filipino creatives in both professional and academic settings. It aims to uncover common usability challenges, evaluate the effectiveness of existing accessibility features, and propose improvements that reflect user needs and promote inclusive design practices. Employing a qualitative, exploratory methodology, the study integrates word cloud analysis and the Diffusion of Innovation theory to interpret user feedback and assess attitudes toward accessible tools such as colorblind-friendly palettes and vision simulators. Results show strong support for built-in, customizable accessibility features, with many users expressing openness to adoption if these tools are integrated seamlessly into their workflows. The study contributes to the discourse on digital inclusion by presenting an ethical design framework that promotes accessibility as a standard, not optional, element in graphic design platforms.
8:15 MySecureMap: a Geolocation-Based Recommender System for Safety and Healthy Environment
Preecha Tangworakitthaworn, Poramet Kaewpradub, Patcharapon Hathaichot and Waritthorn Mahasiripanya (Mahidol University, Thailand)
This research presents the MySecureMap system, a geolocation-based recommender system designed for real-time air quality visualization and safety-focused navigation. The system leverages the Air Quality Index (AQI) data provided by both government and private sources. The proposed mechanism has integrated the Long Short-Term Memory (LSTM) neural network's machine learning technique which is capable of predicting the AQI levels and providing the route recommendations. The ultimate goal of the proposed system aims to protect users from the hazardous air pollutants by automatically identifying the risk zones and suggesting the safe routes which are visualized on the map. This paper discusses the proposed approach and the system architecture, the design and the development of the proposed MySecureMap system, and the performance evaluation, highlighting its usability and effectiveness in enhancing the environmental awareness and public health. The performance evaluation results shown that the proposed system met the acceptable accuracy with Mean absolute percent error (MAPE) of 2.39 % when comparing the predicted forecast value with the real AQI data observed for the next 12 hours.
8:30 Historical Ice Velocity Mapping Techniques: Assessing Long-Term Stability of the Prince Harald System
Liang Tang, Gang Qiao, Rongxing Li, Guojun Li and Shi Li (Tongji University, China)
Long-term ice dynamics of the Prince Harald system in Lützow-Holm Bay (LHB), East Antarctica, particularly before 1990, remain poorly investigated, hindering assessments of its historical contribution to regional mass balance. Here, we reconstructed high-resolution ice velocity fields (1973-1989) for its main components - Prince Harald 1, Prince Harald 2, and Prince Harald 3 - leveraging historical Landsat imagery with a robust photogrammetric method, a systematic framework that includes geometric correction, orthorectification, and hierarchical feature matching. Our integrated analysis, combining these velocities with assessments of basal melt, surface elevation change, ice front dynamics, and Passive Shelf Ice (PSI) behavior, reveals a stable state of the system. We find that this stability was characterized by consistent flow velocities, persistent surface thickening, minimal basal melt, and calving events predominantly within the PSI zone. These results indicate an overall trend of stability and net mass accumulation for the Prince Harald system during this period.
8:45 IPBlocks: a Blockchain Ecosystem for Secure IP Registration and Decentralized Marketplace
Sadia Ahmmed, S M Jishanul Islam and Sahid Hossain Mustakim (United International University, Bangladesh); Ridwan Arefin Islam (McGill University, Canada); Subangkar Karmaker Shanto (Purdue University, USA); Salekul Islam (North South University, Bangladesh)
Intellectual properties (IPs) are crucial intangible assets in today's innovator economy. However, their management suffers from bureaucratic inefficiencies and vulnerability to mismanagement. We introduce IPBlocks, a blockchain-based solution that streamlines IP applications, trading, and royalty transfers through a decentralized marketplace. We develop a set of algorithms that allow users to apply for, publish, auction, and transfer IPs. These processes are designed to ensure enhanced security and transparency. The algorithms are implemented through Solidity on the Ethereum platform. These algorithms exist in a smart contract that is deployed on a blockchain network to interface with a user-friendly web application. Large files associated with IPs are stored in a decentralized file storage system. Our performance analysis demonstrates the system's feasibility and scalability for large user bases, testing the performance in local and deployed settings. IPBlocks aims to revolutionize IP ownership and management by reducing processing times and improving royalty distribution. We open-source our codebase to the community to promote collaborative development and transparency.
9:00 Assessing Migraine Severity in University Students Using Low-Channel Wearable EEG
Tan Lee Fan, Huei Qhi Teoh, Hou Kit Mun and Kiruthika Selvakumar (Universiti Tunku Abdul Rahman, Malaysia)
Migraine is a prevalent neurological disorder characterized by episodic headaches and altered cortical activity, often impacting quality of life. Although advanced neuroimaging has enhanced understanding of migraine-related brain changes, these methods are often costly and inaccessible for routine use. This study investigates the feasibility of using a low-channel, wearable electroencephalography (EEG) device to assess migraine severity in university students. Fifty-three participants were stratified by migraine-related disability using the Migraine Disability Assessment (MIDAS). Resting-state EEG data were recorded with the Muse 2 headband and processed via EEGLAB in MATLAB. Power spectral density (PSD) analysis focused on alpha (8-12 Hz) and beta (13-30 Hz) bands at frontal and temporal sites. Participants also completed the Migraine-Specific Quality of Life Questionnaire (MSQv2.1). Although no statistically significant differences were found across MIDAS groups, higher disability levels showed a trend toward reduced alpha and beta power, suggesting cortical hypoactivation consistent with prior migraine research. Quality of life scores did not significantly differ among groups, indicating that migraine severity may not directly correspond to self-perceived well-being. These findings align with previous studies and support the potential utility of consumer-grade wearable EEG devices for non-invasive assessment of migraine severity.
9:15 Characterization of a Solar Rooftop Greenhouse Drying System for Food Applications
Lew Andrew Tria (University of the Philippines - Diliman, Philippines); Kyle Lemuel S Juliano (University of the Philippines Diliman, Philippines)
In the Philippines, especially in urban areas like Quezon City, food waste has become a significant concern. Agricultural innovations such as solar greenhouse dryers have been adapted to address this issue. On the other hand, installation of grid-tied photovoltaic systems in building rooftops in urban areas have been increasing. Merging these two concepts to design an urban solar rooftop greenhouse dryer is proposed to address the two issues of energy production and food wastage mitigation. This work investigated the drying capability of an urban rooftop solar greenhouse drying system and its effect on the output power of solar modules. Results showed that the proposed system is technically viable achieving an average temperature and humidity reduction of 4.25°C and 1.75%, respectively, resulting to average crops weight loss of 90.85%, relatively similar to a regular greenhouse dryer. In addition, energy generation of solar modules in the proposed setup decreased by 9.6% only with respect to a usual grid-tie system.
Computing & Computational Intelligence (CCI) 3
Track C6F8 CCI 6.3: Computing & Computational Intelligence (CCI) 6.3
Room: F8. 507 Monsopiad
Chairs: Mohamad Yusoff Alias (Multimedia University, Malaysia)
8:00 Length-of-Stay Prediction with Data Fusion and Masked Language Modeling
Hieu Ngo (Ho Chi Minh City University of Technology, VNUHCMC, Vietnam); Chau Vo (Ho Chi Minh City University of Technology, Vietnam)
Length of Stay (LOS) prediction plays a pivotal role in efficient healthcare resource management, optimization of care quality, and reduction of treatment costs. One of the key challenges in this domain is the complexity of processing unstructured long clinical notes containing sensitive information. Moreover, limitations persist in effectively processing long texts and integrating multiple information sources. This paper therefore proposes a method integrating two medical data sources: structured demographic and admission data, unstructured long clinical notes. Both can be then transformed into a unified text input. This data combination simultaneously leverages quantitative information from tabular data (demographic and initial admission information) and rich supplementary information from clinical notes (medical history, detailed symptoms, etc.), providing a more comprehensive picture of patient condition and improving prediction performance. Additionally, the paper proposes applying masked language modeling to enhance the ModernBERT model to create more effective general representations of entire patient texts compared to traditional approaches that rely solely on BERT. Evaluated on MIMIC-III data, our resulting Ad-ModernBERT model can yield better LOS predictions consistently in many experiments.
8:15 Real Time Sign Language Recognition Using MediaPipe and Random Forest Classifier
Ajay Kumar Naik G, Imon Kumar Biswas and Venkata Rama Rao Chevula (National Institute of Technology Warangal, India)
Deaf individuals typically use Sign Language (SL), which involves the use of hand and finger motions, to engage in communication with others. However, when interacting with hearing individuals who may not be familiar with SL, there is often a notable challenge in translating these gestures. To address this communication barrier, various research efforts have focused on employing computer vision techniques and sensor data for recognizing hand movements. This paper presents the development of a Sign Language Recognition (SLR) system utilizing a computer vision-based approach. Here, the data is represented in the form of coordinates extracted using Google MediaPipe from individual frames of video. The extracted features through this representation are fed to the Classifier. Here, the random forest machine learning algorithm is used as a classifier. The experimental results reveal that the Sign Language Recognition (SLR) system using Random Forest Classifier (RFC) achieved an accuracy of 99.777%. This performance is superior to that of other considered machine learning algorithms.
8:30 Disease Classification and Report Generation from Chest X-Ray Images Using SigLIP and LLaVA
Kamalnath M S, Harsh Kumar and Jiji Victor Charangatt (Shiv Nadar University Chennai, India)
In real-world scenarios, radiologists often face the challenge of interpreting numerous medical images, emphasizing the need for scalable and effective automated medical image report generation systems. The goal of Artificial Intelligence (AI)-based report generation is to produce clinically accurate and coherent descriptions from medical images, which can help alleviate the workload involved in traditional radiology reporting. In this paper, we present a novel methodology for multi-label disease classification and automated report generation from chest X-ray images. As the first step, our method utilizes the Sigmoid loss for Language-Image Pre-training (SigLIP) model which utilizes the sigmoid loss to predict relevant medical labels based on image features. These predicted labels guide the report generation process where we employ the Large Language and Vision Assistant (LLaVA) model for generating comprehensive radiology reports, taking advantage of its vision-language capabilities to produce more contextual and descriptive outputs. We assess our unified multi-label classification and report generation framework on the MIMIC-CXR dataset using established evaluation metrics. Experimental results demonstrate that our approach exceeds the performance of many existing state-of-the-art methods on multiple evaluation criteria.
8:45 Powerlist-Based High-Level Programming Model for Parallelism
Jeya Ganesh M R and Anshu S Anand (Indian Institute of Information Technology Allahabad, India)
Designing efficient parallel programs is a non-trivial task that requires one to have an in-depth knowledge of the underlying parallel architecture, memory hierarchy and programming model besides other aspects. This may require considerable efforts, time, and cost. To improve the programmers' productivity, there is a need for parallel abstractions that can simplify the programmers job of first specifying the computation and then optimizing it for performance. Powerlist [1] is one such potential parallel abstraction that is able to exploit both parallelism and recursion at the same time, resulting in succinct representations of computations. In this work, we propose the use of Powerlist as a Domain Specific Language (DSL) and implement it as a Source-to-Source compiler that translates the high-level powerlist specification into sequential and parallel C++ programs. Several new powerlist specifications were also devised for computations and in this process, the powerlist notation was also enriched. The proposed framework considerably improves the programmers' productivity by providing a high-level abstraction resulting in succinct and intuitive specification of computations, absolving them of the task of exploiting parallelism and tuning for performance, which is taken care of by the implementation.
9:00 AI Detection and Watermarking: an Advanced Deep Learning Framework for Real-Time Content Authentication and Digital Fraud Prevention
Aneesh Narayan Bandaru (IIIT Naya Raipur, India); Karthikeya Prachodhan Mudumba (IIIT Naya raipur, India); Mallikharjuna Rao K (IIIT Naya Raipur, India)
The emergence of large language models (LLMs) has transformed the landscape of digital content generation, enabling the creation of human-like text at scale. However, this capability has introduced profound challenges in verifying content authenticity and combating misinformation. As AI-generated content becomes increasingly indistinguishable from human-authored text, existing detection tools struggle to maintain accuracy, especially when confronted with adversarially paraphrased or subtly modified outputs. In response to these challenges, this study proposes a dual-layered framework that combines a CNN-LSTM hybrid deep learning model for AI content detection with robust watermarking mechanisms for authentication. The detection model is trained on a rich dataset comprising human and AI-generated sentences across diverse linguistic patterns. Simultaneously, watermarking techniques, token-level, logit-wise, and frequency-based, are applied to embed statistical fingerprints in AI-generated text to support traceability. Evaluations demonstrate high detection accuracy, reduced false positives, and watermark resilience even under adversarial transformations. The proposed approach contributes toward building a reliable and scalable solution for ethical AI content regulation and digital traceability.
9:15 gptPromptFuzz: LLM Prompt Engineering-Based Seed Generation for Effective Fuzzing
Darshan Lohiya (National Institute of Technology Warangal, India); Vivek Yelleti (National Institute of Technology Warangal, India & Institute for Developement and Research in Banking Technology, India); Sangharatna Godboley (National Institute of Technology Warangal, India); Pisipati Radha Krishna (National Institute of Technology, Warangal, India)
Fuzz testing is one of the popular techniques for evaluating software reliability. Its effectiveness largely depends on the quality and diversity of the initial seed inputs. Traditionally, these seeds are generated randomly, which may limit the effectiveness of the fuzzing process. However, generating seeds based on an analysis of the target code can significantly improve the performance of these tools. To address this, we proposed a Large Language Model (LLM)-based seed generation approach for effective fuzzing and named it gptPromptFuzz. In our approach, initially, one meta-prompt is designed in accordance with the objective of diverse seed generation. To further enhance diversity, we construct ten additional prompts that are semantically equivalent to the meta-prompt. Each of these prompts is independently processed by LLM to produce unique seeds. The experimental results demonstrated that the proposed gptPromptFuzz outperformed random AFL in generating seeds effectively, with reduced execution time, in all 45 benchmark C programs. Further, a larger number of paths are obtained in 41 out of 45 programs.