Optimizing Brain Signal Classification through Computational Neuroscience with EMOTIV and Free Software

This study presents a method for the acquisition, preprocessing, and classification of brain signals aimed at brain mapping and the development of a brain-computer interface (BCI) using non-invasive electroencephalography (EEG). The proposed system is based on the hypothesis that imagined motor actions can be translated into control commands to displace a robotic prototype in four basic directions: right, left, push, and pull. For this purpose, raw biosignals were collected with EMOTIV headsets (EPOC X and Insight), exported in .csv and .edf formats, and subjected to a preprocessing pipeline including artifact removal, digital filtering (1–40 Hz), and statistical outlier detection to ensure data quality. The signals were then analyzed with different classification methods (J48 Decision Tree, RandomForest, NaiveBayes, and Support Vector Machines) to identify the most suitable model for BCI-based neurocontrol. Results show that RandomForest achieved the highest accuracy, reaching 100% with external test data, followed closely by J48 with 99.01%, while NaiveBayes and SVM demonstrated limited reliability under cross-validation. Finally, the methodology was validated through the neurocontrol of a robotic prototype using Arduino hardware, where participants successfully executed directional commands via mental activity. These findings demonstrate the feasibility of integrating consumer-grade EEG devices, free software, and machine learning techniques to develop robust and accessible BCI systems for assistive robotics. The contribution of this work lies in optimizing mental command recognition and highlighting the potential of EEG-based neurocontrol technologies to enhance mobility and autonomy in individuals with motor disabilities. Smart citations: https://scite.ai/reports/10.61467/2007.1558.2026.v17i2.1264Dimensions.Open Alex.