Determination of Increased Mental Workload Condition From EEG by the Use of Classification Techniques

DURU, Adil Deniz · 2019 · Crossref

DOI: 10.7240/jeps.459420

archive: archived pipeline: cataloged

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the challenge of real-time monitoring of human mental states by developing a classification framework to distinguish between eyes-open (EO), eyes-closed (EC), and increased mental workload (MW) conditions using electroencephalogram (EEG) data. The motivation stems from the need for rapid, non-invasive assessment of cognitive status, particularly for brain-computer interface applications where decision latency must be on the order of one second. The research aims to determine if short-duration EEG epochs can effectively differentiate these states using machine learning techniques. The experimental design involved 15 volunteers who underwent three distinct conditions: three minutes of resting with eyes open, three minutes of resting with eyes closed, and a mental arithmetic task designed to increase workload. 16-channel EEG data were recorded at a 1 kHz sampling rate and bandpass filtered between 0.5 Hz and 40 Hz. The data were segmented into one-second epochs, and power spectral features were extracted for delta, theta, alpha, and beta bands across all channels. These features were concatenated to form a 64-dimensional input vector for each epoch. After removing artifacts related to eye movements, the pooled data from all subjects were split into 60% training and 40% testing sets. Three supervised classification algorithms—K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Decision Tree—were implemented to classify the epochs. The results demonstrated that KNN outperformed the other methods, achieving a classification accuracy of 0.94. In comparison, SVM and Decision Tree methods yielded accuracies of 0.88 and 0.89, respectively. Detailed performance metrics indicated that KNN also achieved superior precision, recall, and F1-scores across all three classes. The Decision Tree analysis revealed that beta band power values from specific channels (P3, T7, T8, F3), alpha band power from the right occipital region, and delta band power from C4 were the most discriminative features in the initial layers of the tree. The study confirmed that pooling data across subjects allows for the generalization of the model, enabling the classification of new subjects without requiring individual calibration. The significance of this work lies in its demonstration that high classification accuracy can be achieved using short, one-second EEG epochs, which is critical for real-time applications. The findings suggest that power spectral features from standard frequency bands are sufficient to distinguish resting states from increased cognitive load. The superior performance of KNN over SVM and Decision Trees in this context highlights the utility of nearest-neighbor methods for EEG-based state recognition when inter-subject variability is managed through data pooling. This approach supports the development of efficient brain-computer interfaces capable of monitoring mental workload with minimal training data and sensor requirements.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.8-27b-gittensor on 2026-08-23; verification: pending re-verification.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• neuro workload indices
• drowsiness detection algorithms

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: physiological data