Scalogram-Based Blink Detection in EEG with GoogLeNet, Gaussian Naive Bayes and Majority Voting Using meBaL Dataset

Low mobility is a significant global issue, affecting over 16% of the world's population, or one in every six persons. It impacts the quality of life and daily routines of those affected. To address this issue, we initiated a project to command a robotic wheelchair using the NeuroSky MindWave headset. The first step was to develop an eye-blink detection algorithm using the meBaL open dataset which is widely used to analyze eye-blink behavior and measure the user's level of attention. However, the main challenge in using only EEG data is distinguishing between blink and non-blink states. Accurate detection of blink behavior in EEG data is crucial for understanding how the brain reacts to different situations. This understanding can provide valuable insights for developing assistive technologies that adapt to the user's cognitive state, improving their quality of life. This paper proposes a hybrid approach for recognizing eye blinks in EEG data. In this study we utilized EEG signal scalograms-a time-frequency representation, combined with Gaussian Naive Bayes for classification with majority voting and GoogLeNet for feature extraction, achieving an impressive accuracy of 95.15%. This research aims to contribute to the growing field of neurophysiological studies and pave the way for future innovations, particularly assistive technologies for people with limited mobility.