Abstract
Brain-computer interfaces based on motor imagery tasks require high classification accuracy of electroencephalographic signals, which is a key condition for effective system performance. However, inter-subject variability and sensitivity of the signals to noise significantly complicate their processing. The aim of this study was to develop and evaluate the effectiveness of subject-specific optimization of the electroencepha-lographic signal processing pipeline for multiclass motor imagery classification.A comprehensive approach was applied, including signal segmentation, frequency and spatial filtering, feature extraction, and classification. A comparison between baseline and optimized configurations was performed using cross-validation.The results showed that individual parameter tuning improves classification accuracy. The best performance was achieved using logistic regression, with an accuracy of 86.5% and a Cohen’s kappa coefficient of 0.82. The results confirm the effectiveness of the proposed approach and its potential for developing reliable control systems.