Detecting Mental Arithmetic Activities Using the Determinant Feature of EEG Signals in the Dynamic Graph Mode Decomposition Domain
The analysis of electroencephalogram (EEG) signals recorded under mental workload supports the identification of learning disorders. In addition, the assessment of mental arithmetic performance is relevant to the design of brain–computer interface (BCI) systems. This paper applies dynamic graph mode decomposition (DGMD), a recently introduced digital signal processing technique, to automatically decompose multichannel EEG data into intrinsic mode functions (IMFs), and introduces an optimized matrix determinant of EEG signals as a discriminative feature for mental workload analysis. Each EEG channel is decomposed into IMFs by DGMD, the determinant is computed for the IMFs of that channel, and the mean determinant value is retained as the representative feature of the channel. Optimization algorithms are then used to select the most informative channels, and the features extracted from the selected channels are supplied to the classification models. Several optimization algorithms are compared in order to recommend the group of channels that contributes most to the mental workload task. Extensive experiments are conducted to identify the brain lobe that is most informative for mental workload and mental arithmetic performance. Two classification models are proposed for these two problems and are evaluated on the Zyma dataset, which defines two classification tasks: mental counting versus mental resting and good counter versus bad counter, representing mental arithmetic workload and mental arithmetic performance, respectively. Accuracies of 99.24% and 100% are obtained for the two tasks when all 19 channels are used, and of 99.24% and 99.77% when 14 and 13 channels selected by particle swarm optimization are used. A ten-fold cross-validation strategy is used to obtain robust estimates. The results further indicate that the frontal lobe is the most informative region for both mental arithmetic workload and performance, whereas the central lobe contributes comparatively little to the identification of mental arithmetic activities. The proposed models may assist physicians and neurologists in detecting learning disorders and in designing effective BCI systems.
Authors
- Wael Korani (ORCID: https://orcid.org/0009-0008-7663-1200)
- Hesam Akbari (ORCID: https://orcid.org/0000-0002-9746-4981)
Institutions
- University of North Texas (US)
Publication Details
- Journal
- Mathematical and Computational Biology
- Published
- 2026-11-20
- DOI
- https://doi.org/10.3390/mcb1010003
- Primary Topic
- EEG and Brain-Computer Interfaces
- Type
- article
- Field-Weighted Citation Impact
- 0.00