ECTI Transactions on Electrical Engineering, Electronics, and Communications

Feature and Window Optimization for SNR Estimation in Power-Line-Contaminated Facial EMG Signals

Pornchai Phukpattaranont — 2026-06-29

Electromyography (EMG) signals are widely used in biomedical and rehabilitation applications; however, they frequently suffer from power-line interference (PLI), which compromises signal quality and impacts subsequent analysis. Therefore, it is important to be able to accurately estimate a signal-to-noise ratio (SNR) to assess data quality and guide the method of noise removal. This study investigates the effects of feature number and window size on the accuracy of SNR estimation in the EMG signal contaminated by the PLI signal. The EMG signal is contaminated with synthetic 50 Hz interference at controlled noise levels. Eight features are derived from windows of differing lengths, and the SNR was estimated utilizing 23 regression-based models. Results indicate that waveform activity (WA) and kurtosis (KURT) were found to be the best for estimating SNR. Using window sizes between 250 ms and 2000 ms, these features produced RMSE values from 3.95 to 3.17, demonstrating that larger windows enhance estimation accuracy. In addition, the application of the proposed model is preliminarily validated with real-world facial EMG signals.

Optimal Hyperparameters Solutions for DNN to Improve Accuracy of Indoor Positioning Systems Based on Channel Impulse Response

Thi-Kieu-Lan Ta — 2026-06-29

This paper introduces a fault-tolerant indoor robot localization framework that leverages deep neural networks (DNNs) to process channel impulse response (CIR) data from multiple access points (APs). Unlike conventional CIR-based localization methods that assume stable infrastructure, this work explicitly addresses the challenge of AP failure—an issue largely overlooked in prior studies. Two complementary DNN architectures are investigated: a single-input model that concatenates CIRs from all APs, and a modular multi-input model in which each branch processes the CIR of an individual AP. To improve robustness, we propose a hybrid Hyperband–Simulated Annealing (SA) optimization strategy for both hyperparameters and network parameters. Experiments show that while the single-input network achieves high accuracy under ideal conditions (ADE = 0.552 m), its performance deteriorates significantly under AP outages (up to 1.019 m). In contrast, the multi-input architecture—with branch-specific hyperparameter tuning—consistently maintains strong performance across all failure scenarios, achieving ADEs of 1.15--1.19 m. These results demonstrate the novelty and effectiveness of combining architectural modularity with adaptive optimization to achieve resilient CIR-based localization in dynamic indoor environments.

Optimal Distributed Generation Planning for Loss Reduction and Phase Balancing in Unbalanced Distribution Networks Using the INFO Algorithm

Trieu Ton Ngoc — 2026-06-29

This paper introduces an optimization framework for distributed generation (DG) planning in three phase unbalanced distribution networks (DN), aiming to jointly minimize active power losses and phase imbalance. Unlike conventional methods that assume balanced loading, the proposed model explicitly addresses phase asymmetry and single phase DG integration, common in real world low and medium voltage systems. A composite objective function is formulated, combining power loss with a normalized phase imbalance index under voltage and capacity constraints. The Weighted Mean of Vectors (INFO) algorithm, a recent metaheuristic with strong convergence properties, is applied to solve the mixed integer nonlinear problem. Simulation studies on modified IEEE 33 bus and 69-bus DN with realistic unbalanced loads show that INFO outperforms GWO, MPA, and AOA in loss reduction, phase balancing, and computational efficiency. These findings underscore the relevance of phase aware DG planning in enhancing the resilience and efficiency of modern distribution networks.

Embedded Partial Discharge Classification System Using Transformer Neural Networks on Raspberry Pi

Theerayod Wiangtong — 2026-06-29

Partial Discharge (PD) represents a critical indicator of insulation deterioration in high-voltage equipment. This paper presents an embedded PD classification system using a Raspberry Pi platform with transformer-based neural network architecture. The system classifies four PD types—corona, surface, internal, and noise—using phase-encoded signal data based on IEC 60270 standards. PRPD patterns are converted to sequential vectors through phase encoding: each discharge pulse is mapped to a discretized phase index with its associated charge value. The Raspberry Pi platform integrates an SSD1306 OLED display (128×64 pixels) and five LED indicators for real-time visual feedback. Experimental results demonstrate that the transformer architecture achieves 94.8% validation accuracy compared to CNN baselines at 88.3%. The system enables portable, low-cost condition monitoring without expensive laboratory equipment, demonstrating the viability of embedded artificial intelligence for industrial diagnostic applications.

A Simplified Analog Implementation of Cyclic Shift Chirp Encoding and Decoding for LoRa Communications

Nopparut Saelim — 2026-06-29

Currently, LoRa wireless communication technology has become widely adopted in IoT systems due to its long-range capability and low power consumption. However, LoRa is a technology developed by Semtech, which does not disclose the details of the Cyclic-Shift Chirp encoding process, a core component of LoRa signals. This lack of transparency prevents users from accessing the physical-layer structure of the signal or freely customizing key parameters such as bandwidth and spreading factor. Although such customization can enhance system flexibility, there is currently no officially disclosed method to achieve it. This research proposes a Cyclic-Shift Chirp encoder/decoder circuit built from basic analog components, including adders, subtractors, comparators, and ramp generators, based on a PWM modulation principle. This approach enables researchers and developers to generate LoRa-like signals independently and customize various parameters without introducing limitations. Moreover, the proposed circuit is simple, low-cost, and easy to understand, making it suitable for advanced research, educational experiments, and the design of communication systems that require high flexibility at the LoRa PHY layer.

Adaptive Gamma Weighted Tri Histogram Equalisation

Shubhi kansal — 2026-06-29

This paper introduces a novel image enhancement technique called Adaptive Gamma Weighted Tri Histogram Equalisation (AGWTHE), designed to improve
the visual quality of both grayscale and color images. The proposed method incorporates adaptive gamma correction into a tri histogram equalisation framework, enhancing image contrast while preserving brightness and details. Initially, the image histogram is adaptively clipped using a gamma weighted function to prevent over enhancement and detail loss. The resulting histogram
is then partitioned into three sub histograms based on statistical features such as mean and standard deviation. Each sub histogram undergoes individual histogram
equalisation, followed by a fusion process to generate the final enhanced image. Experimental evaluations demonstrate that AGWTHE effectively enhances image
quality while maintaining natural appearance. The method outperforms several state of the art techniques in both subjective and objective assessments, as validated by metrics including Entropy, FSIM, VSI, and GMSD. The proposed approach is robust, adaptive, and suitable for a wide range of image types and applications.

Efficiency Evaluation of Boost and LLC Resonant Converters for Solar Water Pumping Applications

Nagma Bee — 2026-06-29

This paper investigates the feasibility of employing an LLC resonant converter in a high-efficiency Solar Water Pumping System (SWPS) as an alternative to the conventional boost converter stage. The proposed system integrates maximum power point tracking (MPPT), PI-based DC-link regulation, and V/f control of an induction motor to maintain optimal performance under varying solar irradiance conditions. A detailed MATLAB/Simulink model was developed and experimentally validated using an OPAL-RT 4510 hardware-in-the-loop (HIL) platform. The LLC converter operates at 10 kHz, while the motor frequency is varied between 0–50 Hz to track irradiance-dependent reference speeds. Stagewise power flow analysis was conducted at irradiance levels of 1000, 500, and 300 W/m², comparing both boost and LLC topologies. Results indicate that the motor effectively follows the reference speed without overcurrent, and the LLC-based configuration achieves soft switching with a modest efficiency improvement from 85.6% to 87.0% over the conventional boost design. While the LLC approach may involve higher component costs, the study demonstrates its potential to enhance energy conversion efficiency, reduce switching losses, and improve operational reliability, providing insights into next-generation solar pumping solutions.

Multiclass Non-Technical Loss Detection with Reduced Synthetic Data via a Two-Stage Hierarchical 1D-CNN

Katsarin Seepromting — 2026-06-29

Non-Technical Loss (NTL) detection is a critical challenge in power distribution systems, particularly in regions dominated by conventional mechanical meters and severe class imbalance, as NTL causes substantial revenue loss and undermines system reliability. This paper proposes a two-stage hierarchical 1D-CNN framework for multiclass NTL classification using real-world monthly consumption data from the Provincial Electricity Authority (PEA) in Khon Kaen, Thailand, where over 90% of meters are manually read and NTL rates are the highest in Northeast Thailand. The first stage performs three-class classification (Normal, Energy Theft, Defective Meter) directly on raw imbalanced data, achieving 100% recall for defective meters without synthetic augmentation. The second stage applies Synthetic Minority Over-sampling Technique (SMOTE) selectively to the filtered theft class, reducing synthetic data volume by approximately 50% - thus lowering computational cost -while preserving 99.95% theft recall. Evaluated across four experiments against SVM, Random Forest, XGBoost, and LSTM baselines, the proposed method eliminates manual triage by enabling direct operational routing: repair dispatch for defective meters and legal investigation for energy theft. By minimizing artifact bias and overhead, this work bridges the gap between academic binary models and real-world utility requirements, delivering a robust, deployable solution for NTL management in resource-constrained environments.

Short-Term Load Forecasting for Hospitals Using Transformer-Based Deep Learning with Temporal Memory and Statistics Features

Piyapong Boonsompan — 2026-06-29

Reliable electricity is vital for hospitals, where even short disruptions can affect patient safety and daily operations. This paper presents a single-hospital case study on short-term load forecasting using real half-hourly data from Mae Moh Hospital in Thailand, covering October 2023 to February 2025. To avoid future data leakage and improve model robustness, we design a leakage-aware feature-engineering pipeline that integrates time-of-day cycles, operational context (weekend, holiday, working hour), and temporal memory with statistical features. A 48-step input window (24 hours) is used to predict the next two steps (+30, +60 minutes). We evaluate several compact Transformer-based hybrids (Transformer+LSTM, Transformer+GRU, Transformer+CNN) against baseline methods. Results show that the Transformer+LSTM achieves the best performance (RMSE = 0.45426, MAE = 0.26388, R2 = 0.92571), capturing both daily cycles and fine-grained transitions in hospital demand. The proposed work flow is reproducible and practical for real-world use, enabling hospitals to improve energy planning and prepare for integration into future Virtual Power Plant (VPP) programs.

Aspect-Level Sentiment Analysis Using WangchanBERTa for Fine-Grained Service Insight Extraction in Hotel Reviews

Thanachok Suwan — 2026-06-29

Online customer reviews on hotel booking platforms significantly influence consumer decisions and the reputation of SME hotels in Thailand. The critical challenge for hoteliers lies not in data scarcity but in efficiently extracting actionable insights from unstructured Thai-language reviews. Thus, this research presents an automated sentiment analysis and strategic insight generation system for hotels, leveraging WangchanBERTa, a Thai-specific deep learning model. The system comprises two phases. Phase 1 develops a sentiment classification model using 10,040 Thai hotel reviews collected from Agoda, Booking.com, Traveloka, and Trip.com, categorizing sentiments into positive and negative classes while identifying key topics such as pricing, service quality, and cleanliness. Phase 2 extracts granular aspect-level insights across 11 service dimensions to detect nuanced patterns, such as customers being satisfied with service but dissatisfied with pricing. Experimental results demonstrate robust model performance with an accuracy of 92.62% and a macro F1-score of 94.86% for overall sentiment classification. For aspect-based sentiment analysis, the system achieved 98.66% accuracy with macro precision of 93.62%, recall of 94.62%, and F1-score of 94.22%, alongside effective real-world insight extraction capabilities. This framework enables hoteliers to deeply understand customer voices, transform data into actionable business strategies, and enhance competitive positioning in Thailand's tourism sector.

A A Fully Balanced Bandpass Filter and Its Application

Yotaka Tungtragul — 2026-06-29

This article discusses the design of a fully balanced bandpass filter and its application. The first circuit is a fully balanced bandpass filter composed of six NPN transistors and two pairs of capacitors and resistors. This fully balanced topology inherently offers good common-mode rejection and utilizes current tuning for frequency control. The pole frequency can be changed by tuning the bias current. This circuit design is relatively simple and compact. However, it does exhibit a temperature sensitivity flaw that affects both the pole frequency and the overall harmonic distortion, which is relatively high. The high sensitivity is primarily due to the temperature dependence of the transistor's transconductance. Therefore, the second circuit redesign is an improvement to rectify the drawbacks of the first circuit using the CAPRIO technique, aimed at reducing overall harmonic distortion and reducing the impact of temperature on the transistor characteristics. This circuit consists of four NPN transistors, three resistors, and two capacitors. The core improvement involves replacing the temperature-sensitive active emitter resistance (r_e) with stable passive resistors, thereby fixing the pole frequency's dependence on temperature. This use of the CAPRIO method allows the circuit to achieve better linearity and temperature stability. Although the pole frequency adjustment of this circuit changes through modifications in the resistor and capacitor values, sacrificing electronic tuning for enhanced stability, the outcomes of the second circuit significantly enhance performance: Total Harmonic Distortion (THD) is reduced tenfold (from 1.12% to 0.11%), and the pole frequency exhibits negligible temperature drift (741.31 kHz across 0°C to 100 °C) compared to the first circuit. reduce the temperature sensitivity of the pole frequency and reduce the overall harmonic distortion of the signal. Finally, there is an application for utilizing the bandpass filter as a component of an oscillator circuit. The result achieved is the capability to create sinusoidal oscillators using the proposed circuit.

Event-Driven Neuromorphic Processing for Smart Building Sensor Networks

Luigi Carlo De Jesus — 2026-06-29

This study presents an event-driven neuromorphic framework for intelligent indoor air quality (IAQ) monitoring in smart building environments. The proposed system employs biologically inspired spiking computation using the Leaky Integrate-and-Fire (LIF) neuron model integrated with Spike-Timing-Dependent Plasticity (STDP) learning for adaptive environmental processing. Real-world IAQ parameters—including particulate matter (PM₁, PM₂.₅, PM₁₀), carbon dioxide (CO₂), carbon monoxide (CO), total volatile organic compounds (TVOC), and ozone (O₃)—were acquired from nine building zones and encoded as asynchronous spike events. A fractional-order Kalman filter (FOKF) achieved an average 5.6% noise reduction, stabilizing the signal prior to spike encoding. The neuromorphic model achieved a mean detection accuracy of 10.93%, an average response time of 10.43 steps, and an energy efficiency score of 8.85, reflecting selective sparse firing and low computational overhead. When compared with Long Short-Term Memory (LSTM) and regression models, the neuromorphic system delivered faster response times and significantly higher energy efficiency—nearly nine times lower computational cost—while maintaining comparable event responsiveness. These results demonstrate that event-driven neuromorphic computation offers a scalable, low-power, and adaptive solution for real-time IAQ monitoring in smart building systems.

High-Efficiency Dual-Cascade DC–DC Wide Bandgap Converters Architecture for Tsunami Monitoring

Bagus Irawan — 2026-06-29

Subsea tsunami-meter networks require reliable long-distance power delivery, yet current systems rely on grid-supplied AC feeders that are costly, difficult to deploy in remote regions, and vulnerable during extreme events. AC transmission further introduces reactive losses and reduced efficiency over long subsea cables, motivating a compact, renewable-powered HVDC alternative. This work presents a renewable-driven HVDC architecture combining solar–wind generation, lithium-based battery storage, a dual-cascade high-gain boost converter for long-distance delivery, and a controlled buck stage for regulated sensor-node supply. High-frequency wide-bandgap converters with PI regulation achieve low ripple, stable current control and high efficiency. Hardware results confirm minimal steady-state error and performance comparable to commercial subsea power units. Cable modelling shows that long HVDC links naturally filter ripple while slowing dynamic response. The prototype demonstrates that a high power density, renewable, fully DC system can replace grid-dependent AC infrastructure, reducing cost and enabling scalable, autonomous tsunami-monitoring large networks in remote regions.

Low-Latency Dual-MCU Hardware-in-the-Loop Platform Using Analog-Domain Communication for Electric Drive Applications

Tanpisit Atipasaworn — 2026-06-29

Real-time validation of electric-vehicle (EV) motor-drive controllers remains constrained by the high cost and communication latency of existing hardware-in-the-loop (HIL) systems. To address this limitation, this paper presents a dual-microcontroller HIL platform that enables deterministic, low-latency testing using readily available components. Two Texas Instruments TMS320F28379D digital signal controllers are used to partition the control and plant domains. The first MCU executes cascaded PI-based speed and current regulators, while the second numerically simulates the DC-motor–chopper–vehicle dynamics at a 10 µs step size. A distinguishing feature of the proposed system is its analog-domain signal exchange: the controller’s PWM duty output is low-pass filtered and sampled by the plant MCU, while plant feedback (armature current and speed) is returned through DAC–ADC links. This architecture eliminates protocol overhead inherent in SPI or serial communication and achieves a measured 8.67 µs round-trip latency, ensuring deterministic real-time coupling. Experimental validation using step and mixed-drive-cycle profiles demonstrates tracking performance comparable to a single-MCU benchmark, with reduced current ripple and improved modularity. The entire workflow is implemented through Simulink auto-code generation, requiring no manual driver coding. Beyond providing a cost-effective alternative to commercial HIL simulators, the platform offers a transparent and reproducible framework for research and education in electric-drive control. This contribution highlights how analog-coupled dual-MCU architectures can deliver sub-10 µs responsiveness and structural realism, forming a scalable foundation for next-generation EV HIL development

Temperature-Dependent Hall Effect Analysis and Physics-Based Modeling of Indium Arsenide

Atchariya Phuangyod — 2026-06-29

This paper presents a physics-based numerical framework for modeling Hall-effect transport in n-type indium arsenide (InAs) using the Van der Pauw configuration under sinusoidal electrical excitation. The framework is implemented in a numerical computing environment and captures the coupled relationships among drive current, magnetic flux density, specimen thickness, and the Hall coefficient over a wide range of electrical and thermal operating conditions. The model incorporates temperature-dependent bandgap narrowing via the Varshni relation, effective carrier masses, donor concentration, and mobility behavior following Matthiessen’s rule, together with geometric attributes and operating variables such as magnetic-field strength, temperature ranging from 200 K to 500 K, and time-varying excitation. The computational engine predicts temperature- and frequency-dependent transport quantities, including Hall voltage, carrier concentration, mobility, resistivity, conductivity, and the transit-time-limited cutoff frequency. The resulting framework provides a unified and reproducible computational tool that links microscopic transport physics to macroscopic device-level behavior and supports the analysis and design of InAs-based Hall devices without relying on empirical calibration.

Congestion-Driven Dynamic Hosting Capacity Enhancement Framework using Distribution Network Reconfiguration

Thunpisit Pothinun — 2026-06-29

Traditional hosting capacity (HC) assessment methods for distribution networks commonly rely on static rule-of-thumb criteria, which often impose overly conservative limits leading to the underutilization of existing network assets. However, modern distribution networks operate under high dynamic load demand and power generation. Addressing these dynamics, this study proposes a congestion-driven framework to enhance HC through optimal radial network reconfiguration. An optimization-based network reconfiguration approach was employed to alleviate congestion and increase HC in the distribution network. Furthermore, the framework integrates the calculation of Dynamic Operating Envelopes (DOE) to allocate active power export and import limits for individual distributed generation (DG) and loads. Validated on a modified IEEE 33-bus system, the simulation results demonstrate that the proposed framework extends the allowable DG penetration from 280% in the static base case to 500%, while simultaneously reducing daily operational costs by up to 12.0%.