IEEJ Transactions on Electrical and Electronic Engineering最新文献

Heart rate variability (HRV) has been analyzed to assess the effects of food on autonomic nervous activity, but HRV is reportedly influenced by respiration. Reducing respiratory artifacts in HRV analysis requires respiratory signals. In this study, we estimated respiratory signals from electrocardiogram (ECG), compared these signals with actual respiratory signals, and investigated HRV indices before and after reducing respiratory artifacts based on estimated or actual respiratory signals during functional food intake. ECG and actual respiratory signals were recorded from five healthy males during gamma-aminobutyric acid (GABA) or placebo intake. Estimated respiratory signals were obtained using eight types of ECG-derived respiration (EDR) methods. The highest similarity to the actual signal was observed in the signals estimated from the EDR method based on the fourth-order central moment. Comparing the LF/HF ratios (one of the HRV indices), the difference between GABA and placebo decreased at 60 min post-intake after reducing respiratory artifacts than before reducing them; a decrease of 39.5% using the estimated signals and a decrease of 48.7% using actual signals. Our findings indicate that the EDR method provides the possibility of analyzing HRV with reduction of respiratory artifacts using only ECG, similar to using the actual respiratory signals, in functional food studies. © 2025 The Author(s). IEEJ Transactions on Electrical and Electronic Engineering published by Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

Aiming at the problem of low prediction accuracy of existing models due to the strong intermittency and large fluctuation of the transmission line ice-covering process. Based on the meteorological time series information of icing, a combined forecasting model of icing based on empirical mode decomposition (EMD), kernel principal component analysis (KPCA) and long short-term memory network (LSTM) was proposed. The method firstly uses EMD to decompose the meteorological data, which reduces the instability of the original meteorological series and obtains the modal and residual components with different center frequencies; secondly, the KPCA algorithm is used to extract the main components of the input features and reduce the dimensionality of the input features; and finally, the selected multi-input features are used to build an LSTM network to complete the prediction of the ice cover thickness. The proposed prediction model is validated by example simulation. The results show that the prediction accuracy of the combined EMD-KPCA-LSTM model is further improved compared with other models, which verifies the effectiveness of the proposed method. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

To develop a detection system for abnormal situations, such as accidents, for a person living alone, we investigated a method to update a stochastic model of daily indoor activity sounds (daily sounds). This method adapts to changes in the living environment, updating clusters of daily sounds by the k-means clustering method. Additional daily sounds were periodically incorporated, the clusters were updated, and the parameters of the stochastic model were recalculated. The results indicate the feasibility of continuously updating the stochastic model for long-term daily sounds. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

Parkinson's disease (PD) affects over 10 million individuals globally, making it one of the most common neurological disorders. Despite its prevalence, no definitive cure or therapy exists to halt its progression. PD symptoms majorly influence patients' everyday lives, so prompt identification is essential. Because the disease starts in the brain, we used electroencephalography (EEG) data in our research. We used the publicly accessible dataset ‘EEG: Simon Conflict in Parkinson's’, which consists of EEG recordings from 28 people with PD (ON–OFF medication) and 28 healthy people (CTL). Due to the sensitivity of EEG signals, noise components were removed using Independent Component Analysis (ICA) combined with the IClabel model. Cleaned signals were reconstructed and analyzed into five primary frequency bands: delta, theta, beta, alpha, and gamma with statistical features. Boosting methods from the ensemble algorithm family were applied to evaluate classification performance. The classification results are presented for two labels (healthy individuals and PD) and three (healthy individuals, PD in ON medication, and PD in OFF medication). The XGBoost model achieved the best classification performance, achieving high accuracy, sensitivity, and specificity within a reasonable computation time. The XGBoost model combined with ICA-ICLabel achieved 99.71% accuracy in PD-CTL classification and 91.35% accuracy in three-class classification (CTL-ON-OFF). © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

The traditional model predictive voltage control (MPVC) for grid-forming inverters relies heavily on accurate system parameters, which can impact voltage prediction performance. To address this issue, this paper proposes a data-driven model-free predictive voltage control strategy (DD-MFPVC). Initially, the influence of system parameters on traditional MPVC is analyzed for grid-forming inverters. Then, the data-driven model for grid-forming inverters is established, which is identified using the least-squares method. Furthermore, by applying the multi-layer recursive principle, the future data-driven model is derived from historical data, with recursive coefficients estimated using least squares to mitigate the one-step delay effect in the data-driven model. Subsequently, the voltage vector reference for the inverter is calculated based on deadbeat control principles, integrating space vector modulation to ensure voltage prediction accuracy. The proposed DD-MFPVC allows real-time updating of the data-driven model, eliminating dependency on system accuracy, and reducing prediction errors. The effectiveness of the proposed DD-MFPVC is validated through experimental comparisons. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

This study presents a technique designed to prevent electrostatic discharge during the cleaning of semiconductors using a two-fluid spray composed of pure water. The current generated by spraying pure water with a prototype Faraday cage was measured. Furthermore, a shadow Doppler particle analyzer was used to measure the velocity and size of the airborne droplets, clarifying the relationship between the generated current and droplet characteristics. The results confirmed that the current generated during two-fluid spraying exhibits positive polarity and that the droplet charge increases with droplet velocity. We further demonstrated that the charging of pure-water droplets can be controlled by installing an inductive charging element immediately downstream of the pure-water jet. This phenomenon is believed to occur because the electrons induced by the charging element flow through the continuous fluid region along the nozzle wall toward the ground. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

This study proposes statistical models for mitigating the risks associated with variable power sources by establishing a headroom to absorb prediction errors when PV power plants perform day-ahead planning without the use of battery systems or intraday market procurement. Specifically, three models are proposed in this study. The first model utilized kernel density estimation to analyze the distribution of previous errors and determine error thresholds based on the desired probability set as the headroom. The second and third models utilized different sample sets that adjusted the headroom setting in response to changes in PV power generation. These models generated polynomial surfaces using PV-predicted values and clearness index ($��\mathrm{CI}�$) as features, allowing the error thresholds to vary based on the required probability. Furthermore, separate polynomial surface models were developed by constructing distinct models for both predicted and measured $��\mathrm{CI}�$ values. The optimal model demonstrated only 120 annual negative imbalances at 30-min intervals, with an average headroom to PV generation prediction ratio of 65%, indicating that most of the headroom absorbed prediction errors. This research has significant implications for enabling PV power plants, which require reserve power, to enhance to the stability of the overall power grid while simultaneously serving as a reliable power source themselves. © 2025 The Author(s). IEEJ Transactions on Electrical and Electronic Engineering published by Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

Decoding fundamental human tastes using EEG signals involves examining the brain's electrical activity to better understand how it responds to various flavors, including sweet, sour, salty, and bitter. This method uses electroencephalography (EEG) to capture and understand neural processes relevant to taste perception, revealing how the brain stores sensory information. Understanding the neurological foundation of taste can help medical professionals diagnose and treat taste-related abnormalities caused by aging, trauma, or illnesses like COVID-19, Parkinson's disease, and Alzheimer's disease. This information may enhance product development in the industrial sector, particularly in the food and beverage industry, by tailoring goods to better meet customer preferences based on a deeper understanding of taste reactions. In scientific studies, deciphering brain signals associated with taste experiences is critical for neuroscience research, as it improves our understanding of how the brain processes information. The paper's originality stems from its multidisciplinary approach. It integrates information and methodologies from several domains, including neuroscience, biotechnology, and machine learning, to provide a novel way to decipher brain processes. Deep learning techniques and artificial intelligence are being used to decipher complicated patterns in EEG data, paving the way for practical applications such as automated and customizable taste perception assessment devices. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

Accurate estimation of state-of-charge (SOC) and state-of-health (SOH) is crucial for optimal battery system performance and longevity. To enhance SOC estimation precision, this study proposes a backpropagation neural network-adaptive unscented Kalman filter (BP-AUKF) algorithm for co-estimating SOH and SOC. It is based on a second-order Thevenin equivalent circuit model and the forgetting factor recursive least squares (FFRLS) algorithm. Initially, the FFRLS algorithm determines the model parameters. Subsequently, the BP neural network algorithm estimates SOH as the number of iterations varies. Utilizing the corrected effective battery capacity, the AUKF provides an initial SOC estimate, which the BP neural network algorithm then refines, eliminating estimation errors. The proposed algorithm's superiority is demonstrated through simulations under the US06 highway driving schedule, Beijing Dynamic Stress Testing, Federal Urban Driving Schedule, and constant current conditions. Compared to the AUKF algorithm, it exhibits enhanced SOC estimation accuracy, achieving a mean absolute error below 2% and a root mean square error below 2.1%. Thus, this method ensures high accuracy, strong adaptability, and safe lithium-ion battery operation. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

In practice, impulsive noise may significantly degrade the filtering performance of constrained adaptive filtering (CAF) algorithms derived from the second-order signal statistics. In this paper, two robust constrained arctangent least mean square (CATLMS) algorithms are proposed to overcome this problem, inspired by the boundedness of the arctangent function against outliers. First, a CATLMS algorithm is derived using the gradient descent method. To accelerate the convergence rate in the case of correlated input signals and improve steady-state performance, a recursive CATLMS (RCATLMS) algorithm is further proposed based on the matrix inversion lemma. The computational complexity of our proposed algorithm is comparable to that of other existing robust algorithms. Simulation results demonstrate the effectiveness of our proposed algorithms against impulsive noise environments and the superior filtering performance compared to other CAF algorithms. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.