Mental arithmetic task load recognition using EEG signal and Bayesian optimized K-nearest neighbor

Cognitive load recognition during mental arithmetic activity facilitates to observe and identify the brain’s response towards stress stimulus. As a result, an efficient mental load characterization approach using electroencephalogram (EEG) signal and Bayesian optimized K-Nearest Neighbor (BO-KNN) has been proposed in this work. The study has been conducted on a recorded EEG dataset of 30 healthy subjects who were exposed to an arithmetic questioner. To obtain artifacts free EEG signal, the Savitzky–Golay filtering approach has been utilized. Further, the decomposition of the extracted EEG signal has been carried out using stationary wavelength transform. In this work, the entropy based feature extraction has been performed followed by F-score based feature selection. Top 40 features having the highest precedence have been used for classification using BO-KNN. The rigorous experimental analysis has been performed to analyze the effectiveness of the proposed method over other state-of-the-art methods and it shows that the classification accuracy is substantially improved.