Deep Transfer Learning for Subject-Independent ERP-based BCIs

Kassymzhomart Kunanbayev, Dias Azhigulov, Berdakh Abibullaev, Amin Zollanvari

2021

Closed access Metadata (CSL-JSON)
DOI 10.1109/bci51272.2021.9385311
Data confidence 70 medium
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

Designing subject-independent Brain-Computer Interfaces remains to be an open question for developing systems that can capture the inter-subject intrinsic brain features and classify them with reasonable accuracy. This paper presents the application of the state-of-the-art deep transfer learning architectures on classifying ERP signals. We report 66.87%, 67.64%, 65.58%, and 71.93% test classification accuracy for DenseNet121, DenseNet201, Xception, and VGG-16 models, respectively. The experimental results demonstrate the viability of our approach in subject independent ERP-signals classification and suggest the better performance of models with fewer layers in classifying ERP signals.