A new online event related potential based brain-computer interfaces using an ensemble classifier

Author

Onishi, Akinari ; Natsume, K.

Author_Institution

Dept. of Brain Sci. & Eng., Kyushu Inst. of Technol., Kitakyushu, Japan

fYear

2013

fDate

6-8 Nov. 2013

Firstpage

1033

Lastpage

1036

Abstract

The brain computer interface (BCI) records brain signals, then translates them into commands for control devices for a better quality of life. The event related potential (ERP) has been used to drive the BCIs, but they still have room for improving its performance toward a practical use. In this study we evaluated an online ERP-based BCI system that had an ensemble classifier with overlapped partitioning and a stimulator which provided image intensifications. We also employed and compared four types of visual stimuli: a neutral, a positive, a negative and a traditional gray/white letter stimulus. As a result, ERP-based BCI system with the neutral, the positive and the negative image achieved significantly higher information transfer rate (41.0 ± 2.2 bits per minute at best) than that with the conventional gray/white letter stimulus.

Keywords

brain-computer interfaces; learning (artificial intelligence); medical signal processing; signal classification; visual evoked potentials; brain signals; brain-computer interfaces; ensemble classifier; event related potential; image intensifications; negative stimulus; neutral stimulus; online ERP-based BCI system; overlapped partitioning; positive stimulus; quality of life; traditional gray-white letter stimulus; visual stimuli; Accuracy; Brain-computer interfaces; Databases; Principal component analysis; Training; Training data; Vectors;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Engineering (NER), 2013 6th International IEEE/EMBS Conference on

Conference_Location

San Diego, CA

ISSN

1948-3546

Type

conf

DOI

10.1109/NER.2013.6696113

Filename

6696113