The Role of EEG Signals: SVM Classification of Cognitive Load as a Support for UX Evaluation

Iksan, Ennu Intan; Mardhia, Murein Miksa · 2023 · Crossref

DOI: 10.26714/jichi.v4i1.11198

archive: archived pipeline: cataloged verified

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

Summary

This study investigates the use of electroencephalogram (EEG) signals to classify cognitive load as a metric for evaluating user experience (UX). Cognitive load, defined as the mental effort required to process information, significantly impacts usability; excessive load can disrupt performance. The research aims to determine if EEG-derived brain wave data can objectively distinguish between "burdened" (difficult/confusing) and "not burdened" (easy) states during interaction with a digital application, thereby supporting UX evaluation with physiological data rather than relying solely on subjective reports. The methodology involved 30 respondents aged 21–24 who completed a questionnaire evaluating the WhatsApp application via Google Forms for seven minutes. During this task, participants wore a Neurosky EEG device to record beta waves (13–30 Hz), specifically the attention signal, which correlates with concentration. Data preprocessing included feature extraction using Fast Fourier Transform (FFT) to analyze amplitude and frequency, followed by normalization of peak signals. The study utilized five input variables: average EEG value, health problems, workload scale, rest intensity scale, and technology proficiency scale. The target variable was the cognitive load label (burdened or not burdened), determined by expert labeling based on task scores. The classification was performed using Support Vector Machine (SVM) algorithms, testing four kernel types: linear, polynomial, radial basis function (RBF), and sigmoid. The results demonstrated that the SVM model could classify cognitive load states, with performance varying by kernel type and feature inclusion. In the testing phase, the model using EEG features achieved an accuracy of 75% with the linear kernel (optimized at cost parameter c=1). However, when EEG features were excluded from the model, the linear kernel achieved a higher accuracy of 89%. The confusion matrix analysis indicated that the model correctly identified the majority of "burdened" cases. The study found that while EEG signals provide physiological insight, the inclusion of EEG features in this specific dataset did not yield the highest classification accuracy compared to models relying on other input variables. The linear kernel was identified as the most effective among the tested kernels for this classification task. The significance of this research lies in its contribution to non-invasive, objective methods for UX evaluation. By validating the use of EEG beta waves and SVM classification, the study supports the integration of physiological measures into human-computer interaction research. Although the highest accuracy was achieved without EEG features in this specific experiment, the framework establishes a viable approach for monitoring cognitive load in real-time. This method offers a potential tool for designers to assess user mental workload and interface complexity, moving beyond subjective self-reports to data-driven UX optimization.

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 qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

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

• neuro workload indices