2026-06-302026-06-302026-05-26DOURADO, Guilherme da Costa. Otimização de interfaces cérebro-máquina baseadas em imagética motora por seleção de features e combinação de modelos: compilado de artigos publicados em 2024 e 2025. Orientador: Cleison Daniel Silva. 2026. 32 f. Trabalho de Curso (Bacharelado em Engenharia Elétrica) – Faculdade de Engenharia Elétrica, Campus Universitário de Tucuruí, Universidade Federal do Pará, Tucuruí, 2026. Disponível em: https://bdm.ufpa.br/handle/prefix/9667. Acesso em:.https://bdm.ufpa.br/handle/prefix/9667This undergraduate thesis addresses the study and optimization of Motor Imagery (MI) based Brain-Computer Interface (BCI) systems, focusing on improving the accuracy and robustness of classifiers applied to Electroencephalography (EEG) signals. The work is divided into two complementary studies, both using the public dataset 2a from the IV BCI Competition. In the first study, the impact of feature selection on the performance of MI-BCI systems was investigated. The methodology included sub-band filtering, application of the Common Spatial Pattern (CSP) method for feature vector generation, and feature selection using the Least Absolute Shrinkage and Selection Operator (LASSO) and SelectKBest algorithms, followed by classification with Support Vector Machine (SVM). The results indicated an average gain of 2% in classification accuracy, demonstrating that careful feature selection contributes to reducing redundancies and improving classifier performance. In the second study, ensemble learning techniques such as Bagging and Soft Voting were explored, applied to the combination of personalized models through Bayesian optimization. The analyses showed an average increase of 14.4% in the mean Kappa index (κm), a metric that evaluates the agreement between classifier predictions and true labels, compared to individual models. However, no statistical significance was observed in the values of zκ, indicating that the gains remained sensitive to the individual variability of EEG signals. Together, the two studies constitute a multi-stage optimization approach for BCI systems, contributing to the development of more accurate, interpretable, and adaptable solutions, with potential applications in clinical and assistive contexts.Acesso AbertoInterfaces cérebro-máquinaImagética motoraSeleção de featuresOtimização bayesianaAprendizado em conjuntoBrain-computer interfaceMotor imageryFeature selectionBayesian optimizationEnsemble learningCNPQ::ENGENHARIAS::ENGENHARIA ELETRICACNPQ::ENGENHARIAS::ENGENHARIA BIOMEDICA::BIOENGENHARIA::PROCESSAMENTO DE SINAIS BIOLOGICOSTrabalho de Curso - Graduação - Monografia