Lobe-wise Cognitive Load Detection using Empirical Fourier Decomposition and Optimized Machine learning

Kunamneni ChervithaLAKHAN DEV SHARMA^*

• VIT-AP University, Amaravati, India

Neural activity is stimulated by cognitive load, which is crucial for comprehending how the brain reacts to stimuli or mental strain. This work investigates the viability of using electroencephalogram (EEG) signals to extract and categorize features in order to assess cognitive load. Each lobe's EEG data was recorded over a four-second period, and we used empirical Fourier decomposition (EMFD) to decompose it up into ten intrinsic modes (IMF). The classification performance was optimized by reducing the feature space and extracting entropy based features from IMFs. Our comprehensive study covered both lobe-wise and overall feature classification using optimized ensemble machine learning (OML) classifiers alongside machine learning (ML) classifiers. OML classifiers with EMFD-based features allowed us to refine our approach. The MAT (mental arithmetic task) dataset showed 97.8% accuracy Ac in detecting cognitive load, while the STEW (simultaneous workload) dataset showed 96.4% Ac. Similarly, this paper presents lobe-wise cognitive load detection, which provides lobe-wise information about brain activity during cognitive tasks. We examined MAT's as well as STEW's 5 lobes respectively. With Ac of 97.8% alongside 96.08%, for frontal lobe demonstrated remarkable proficiency in analyzing a range of cognitive tasks. Our method performed better than the most advanced methods currently in use for detecting cognitive load.