Graph Neural Network and SVM-Based Approach for Output Prediction of Three-Phase Matrix Converter
"Intelligent Signal Processing Techniques for Wearable Healthcare Monitoring Systems"
FREQUENCY BASED EEG SIGNAL ANALYSIS FOR MENTAL FATIGUE DETECTION USING WELCHE’S METHOD OF POWER SPECTRAL DENSITY
ECG Signal Classification Using LSTM Networks for Automated Cardiac Diagnosis
Optimization-Driven Adaptive FIR Filter Identification Using LMS, Gradient Descent, and L-BFGS: A Comparative Performance Analysis
Blockchain 3.0: Towards a Secure Ballotcoin Democracy through a Digitized Public Ledger in Developing Countries
Fetal ECG Extraction from Maternal ECG using MATLAB
Brief Introduction to Modular Multilevel Converters and Relative Concepts and Functionalities
Detection of Phase to Phase Faults and Identification of Faulty Phases in Series Capacitor Compensated Six Phase Transmission Line using the Norm of Wavelet Transform
A Novel Approach to Reduce Deafness in Classical Earphones: MUEAR
A novel mathematical ECG signal analysis approach for features extraction using LabVIEW
Filtering of ECG Signal Using Adaptive and Non Adaptive Filters
Application of Polynomial Approximation Techniques for Smoothing ECG Signals
A Novel Approach to Improve the Wind Profiler Doppler Spectra Using Wavelets
Optimization-Driven Adaptive FIR Filter Identification Using LMS, Gradient Descent, and L-BFGS: A Comparative Performance Analysis
This paper presents a novel framework combining Graph Neural Networks (GNN) and Space Vector Modulation (SVM) to model and predict the output waveform and Total Harmonic Distortion (THD) of a three-phase matrix converter. The converter's topology is transformed into a graph structure, enabling spatial and temporal features to be extracted efficiently. SVM modulation is applied to control the switching sequence, and the resultant waveform is used as the target for training the GNN model. The model achieves high accuracy in predicting voltage profiles and THD, demonstrating the capability of ML-augmented converter analysis.
Adaptive Signal Processing (ASP) algorithms have revolutionized remote health monitoring systems through integration into the biomedical device. With the pressure on global healthcare systems to deliver scalable, patient-centric monitoring frameworks, ASP algorithms can help transform how healthcare services are provided now and in the future. The algorithms adapt to different physiological conditions, suppress noise while extracting critical features in real-time and improve the accuracy and reliability of biomedical diagnostics. In this paper, current advancements and applications of ASP within remote biomedical monitoring are presented, and several adaptive filtering techniques (Least Mean Square (LMS), Recursive Least Squares (RLS), Kalman filters) are described. Wearable biosensors, IoT and ASP work in tandem to augment the capabilities of e-health systems to process physiological data continuously and in real time. A number of case studies (including ECG monitoring, EEG-based brain computer interfaces) are described, which demonstrate the practical utility of ASP in diagnosing cardiovascular anomalies, detecting epileptic seizures and monitoring of respiratory irregularities. This paper also assesses algorithmic performance in terms of convergence rate, efficiency via computational cost and signal-to- noise ratio (SNR). These discussions are supported by a literature review including recent results in medical engineering and signal processing. An experimental setup is proposed by the methodology section that uses simulated biomedical signals, the prototyping of hardware using Arduino and MATLAB-based signal analysis. Our results demonstrate significant improvement in noise suppression and anomaly detection over traditional signal processing techniques. The results support the protean promise of ASP in telemedicine and personalized medicine.
Mental fatigue is a cognitive condition that results in a decline in attention, performance, and alertness due to prolonged mental activity. It is particularly critical in safety-sensitive environments, such as aviation, transportation, and healthcare. Early mental fatigue detection is essential to prevent accidents and improve operational efficiency. Electroencephalogram (EEG) is a non-invasive method for monitoring brain activity and detecting cognitive states, including mental fatigue. EEG signals exhibit frequency-specific changes when an individual experiences fatigue, especially in the alpha, beta, and theta bands. This paper proposes a frequency-based EEG signal analysis for mental fatigue detection using the Power Spectral Density (PSD) method. The PSD, computed using Welch's algorithm, estimates the power distribution across frequency bands, highlighting fatigue-related changes in brain activity. Welch’s method enhances accuracy by averaging periodograms over overlapping EEG segments, enabling effective and real-time mental fatigue detection.
Electrocardiogram (ECG) signal analysis plays a vital role in the diagnosis and monitoring of cardiac diseases. Traditional methods rely heavily on manual interpretation, which can be time-consuming and prone to human error. This research explores the application of deep learning techniques, specifically Long Short-Term Memory (LSTM) networks, for automated classification of ECG signals. Using a publicly available ECG dataset, the proposed model undergoes thorough pre-processing, feature extraction, and training to accurately distinguish between different cardiac conditions. The experimental results demonstrate that LSTM networks achieve high accuracy and robustness in ECG classification tasks, outperforming traditional machine learning approaches. This study highlights the potential of deep learning models to enhance automated cardiac diagnostics, with implications for real-time healthcare monitoring systems. Future work suggests further optimization and validation across diverse datasets to improve clinical applicability.
In this paper we describe the comparative study of adaptive finite impulse response (FIR) filter identification through optimization of three typical algorithms: Least Mean Squares (LMS), batch Gradient Descent (GD), and Limited-memory Broyden–Fletcher–Goldfarb–Shanno (L-BFGS) methods. A standard simulation approach is introduced to evaluate convergence speed, estimation capability, frequency-response fidelity, and computation cost in the presence of the same signal, noise, and system conditions. The MSE-based identification problem is shaped in a convex quadratic space, which guarantees a stable evaluation of the first-order as well as quasi-Newton optimization principles. Results suggest that LMS has both low-complexity stochastic updates and the worst steady-state error, GD has smoother descent and higher accuracy through much optimization, and L-BFGS has faster convergence due to effective curvature approximation, whilst also having the most accurate coefficient and frequency-response reconstruction. These quantitative data show remarkable increases in SNR improvement and final MSE for L-BFGS in comparison to the prior mentioned methods. These results illustrate the applicability of quasi-Newton optimization for block-adaptive digital signal processing (DSP) processes with high accuracy, while illuminating the trade-offs associated with real-time and resource-constrained scenarios.
