Comparative Analysis of ML Models for Multi-Class Activity Recognition Using EEG and Motion Sensor Data

Enhancing Brain-Computer Interface (BCI) technology and activity recognition systems relies on accurate classification of multi-class human activities. Recent advancements in ambulatory EEG and motion sensors have opened new opportunities for real-time, context-aware applications. This study aims to leverage these advancements by classifying human activities based on combined EEG and motion sensor data to improve mobility and independence, particularly for individuals with disabilities. Data were collected using a 14-channel EMOTIV EPOCH Electroencephalogram (EEG) helmet under various activities, including eye blinks, eyebrow movements, head shaking, and walking. A comparative evaluation of multiple Machine Learning (ML) models, including Support Vector Machine (SVM), Logistic Regression, Random Forest, and k-nearest Neighbors (KNN) was conducted to identify the most effective model. The Random Forest model attained a maximum accuracy of 99.99%, and the SVM model reached 99%, demonstrating their superior performance in classifying complex activity tasks. The results highlight the efficacy of integrating EEG and motion sensor data with advanced machine learning models for context-aware, real-time BCI applications. These findings are particularly significant for enhancing mobility and independence in individuals with disabilities, paving the way for robust and adaptive activity recognition systems.