Cordes, Lars (2022) Transfer Learning for Motor Imagery Classification in Low-Cost Brain-computer Interface Systems. Master's Thesis / Essay, Artificial Intelligence.
|
Text
mAI_2022_CordesL.pdf Download (17MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (162kB) |
Abstract
Brain-computer interfaces (BCIs) enable people with impaired mobility to interact with their environment. BCIs translate brain activity measured through electroencephalography (EEG) into control commands, such as the movement of a prosthesis. Two challenges to the general adoption of this technology are the high cost of medical-grade EEG devices and the high variability of the EEG signal between different users. This study introduces a BCI based on a low-cost EEG device that uses transfer learning between users to reduce the need for individual calibration. This BCI architecture was tested and evaluated on two data sets collected using low-cost EEG as well as a benchmark data set collected using medical-grade EEG. On all data sets, a transfer learning approach employing Euclidean alignment significantly outperformed a baseline system. However, accuracies for the low-cost EEG data sets were generally lower compared to the medical-grade EEG benchmark data set. We conclude that the use of transfer learning in BCI tasks should be encouraged and that the affordability of low-cost EEG devices might not make up for the decrease in data quality.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Borst, J.P. and Dhali, M.A. |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 30 Nov 2022 08:44 |
| Last Modified: | 30 Nov 2022 08:44 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/29009 |
Actions (login required)
 |
View Item |