文献类型：期刊文献

英文题名：Optimized stimulus presentation patterns for an event-related potential EEG-based brain-computer interface

作者：Jin, Jing[1,2];Allison, Brendan Z.[2];Sellers, Eric W.[3];Brunner, Clemens[2];Horki, Petar[2];Wang, Xingyu[1];Neuper, Christa[2]

机构：[1]E China Univ Sci & Technol, Key Lab Adv Control & Optimizat Chem Proc, Minist Educ, Shanghai 200237, Peoples R China;[2]Graz Univ Technol, Lab Brain Comp Interfaces, Inst Knowledge Discovery, A-8010 Graz, Austria;[3]E Tennessee State Univ, Dept Psychol, Brain Comp Interface Lab, Johnson City, TN 37614 USA

年份：2011

卷号：49

期号：2

起止页码：181

外文期刊名：MEDICAL & BIOLOGICAL ENGINEERING & COMPUTING

收录：;EI(收录号：20111113757805);WOS:【SCI-EXPANDED(收录号:WOS:000286835100006)】；

基金：This research was supported by the Grant National Natural Science Foundation of China, under Grant No. 61074113 and supported by Shanghai Leading Academic Discipline Project, Project Number: B504.

语种：英文

外文关键词：Brain-computer interface; P300 event-related potential; Flash pattern; EEG

摘要：P300 brain-computer interface (BCI) systems typically use a row/column (RC) approach. This article presents a P300 BCI based on a 12 x 7 matrix and new paradigmatic approaches to flashing characters designed to decrease the number of flashes needed to identify a target character. Using an RC presentation, a 12 x 7 matrix requires 19 flashes to present all items twice (12 columns and seven rows) per trial. A 12 x 7 matrix contains 84 elements (characters). To identify a target character in 12 x 7 matrix using the RC pattern, 19 flashes (sub-trials) are necessary. In each flash, the selected characters (one column or one row in the RC pattern) are flashing. We present four new paradigms and compare the performance to the RC paradigm. These paradigms present quasi-random groups of characters using 9, 12, 14 and 16 flashes per trial to identify a target character. The 12-, 14- and 16-flash patterns were optimized so that the same character never flashed twice in succession. We assessed the practical bit rate and classification accuracy of the 9-, 12-, 14-, 16- and RC (19-flash) pattern conditions in an online experiment and with offline simulations. The results indicate that 16-flash pattern is better than other patterns and performance of an online P300 BCI can be significantly improved by selecting the best presentation paradigm for each subject.