The Analysis and Comparison of LMS-Based Filtering Techniques for EEG Signals: Towards Informed Decision Making

Huamani G.L.; Urure B.S.; Colque E.H.; Jesús Talavera S. · SSRG International Journal of Electronics and Communication Engineering (2024) · DOI: 10.14445/23488549/ijece-v11i8p113

Resumen

Electroencephalography (EEG) is essential for observing brain activity. It offers non-invasive, high-resolution insights into neural dynamics. Despite its clinical and research applications, EEG signals are prone to noise from powerline interference, muscle artifacts and environmental sources. This study evaluates adaptive filtering techniques—LMS, NLMS, PNLMS and IPNLMS—for denoising EEG signals. A dataset of 23 EEG recordings contaminated with noise was used. Accelerometer signals served as reference inputs. The algorithms were assessed using Mean Squared Error (MSE) Signal-to-Noise Ratio (SNR) and Pearson Correlation Coefficient. PNLMS was found to be the most effective. It achieved the lowest MSE (0.9193), highest SNR (1.0768) and highest correlation (46.9025%). While PNLMS excels in noise reduction it has computational demands that may limit its use in wearable devices. NLMS offers a practical balance. It balances performance and efficiency. Future work includes hybrid algorithms. Real-time implementations will be addressed. Adaptive parameter tuning will also be covered. These aim to enhance EEG signal processing and its applications in clinical and research environments.

Instituciones: Universidad Nacional de San Agustín de Arequipa