Comparing the performance of stimulation paradigms used in a hybrid BCI speller with a consumer-grade EEG headset | |
| Author | Tan Santativongchai |
| Call Number | AIT Thesis no.DSAI-25-09 |
| Subject(s) | Brain-computer interfaces Computational intelligence |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Data Science and Artificial Intelligence |
| Publisher | Asian Institute of Technology |
| Abstract | This study investigates the novel Subspeller+FERC stimulation paradigm, a newly proposed hybrid design that integrates the sequential frequency grouping strategy of the Subspeller paradigm (Xu et al., 2020) with the frequency enhanced row and column architecture (FERC) described by Bai et al. (2023). Unlike existing approaches, Subspeller+FERC introduces a dual frequency coding scheme in which each target is represented by both a row frequency and a column frequency. This design aims to enrich SSVEP and P300 discriminability, increase target specificity, and potentially enhance hybrid BCI performance. The study examined whether this hybrid design improves classification accuracy or information transfer rate when compared with the standard Subspeller paradigm using a consumer grade eight channel EEG headset. Twenty participants completed a 16-target spelling task using both paradigms in an offline hybrid sSVEP and P300 framework. Accuracy and information transfer rate were computed for each paradigm, and post hoc analyses were performed on the top performing participants. Group level results showed comparable mean accuracies between Subspeller at 45.16 percent and Subspeller+FERC at 46.62 percent. However, Subspeller achieved a considerably higher information transfer rate of 9.08 bits per minute, compared with 5.40 bits per minute for Subspeller+FERC, largely due to the longer stimulation duration required by the dual frequency row and column sequence. The post hoc analysis showed that the highest performers in the Subspeller paradigm reached 82.4 percent accuracy, while the highest performers in Subspeller+FERC reached 80.0 percent. A clear trend was also observed in which participants who performed poorly in the Subspeller paradigm(<60% accuracy) often achieved higher accuracy when using Subspeller+FERC. These findings indicate that although Subspeller+FERC does not surpass the standard Subspeller at the group level. it offers meaningful benefits for individuals with low accuracy in Subspeller. The paradigm is therefore becoming an alternative paradigm for BCI Speller. |
| Year | 2025 |
| Type | Thesis |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Data Science and Artificial Intelligence (DSAI) |
| Chairperson(s) | Chaklam Silpasuwanchai |
| Examination Committee(s) | Attaphongse Taparugssanagorn;Chantri Polprasert |
| Scholarship Donor(s) | Royal Thai Government Fellowship |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2025 |