Skip to main content
. 2012 Apr 4;7(4):e33758. doi: 10.1371/journal.pone.0033758

Table 1. BCI Competition Spelling Accuracies.

R eSVM CNN-1 MCNN-1 SUP OA-SUP OFF-US OFF-US-T ON-US-T OA-US-T ß RE-OA-US-T OA-US
A 5 72 61 61 67 68 46.8 (4.0) 69.0 (0.0) 64.2 (0.9) 66.5 (0.5) 69.0 (0.0) 9.0 (7.4)
10 83 86 82 88 91 89.4 (1.1) 91.0 (0.0) 86.0 (0.0) 87.0 (0.0) 88.0 (0.0) 62.4 (4.1)
15 97 97 97 96 95 95.8 (1.3) 96.0 (0.0) 94.0 (0.0) 96.0 (0.0) 96.0 (0.0) 86.6 (1.6)
B 5 75 79 77 84 84 76.3 (1.6) 79.0 (0.0) 75.0 (0.0) 75.0 (0.0) 79.0 (0.0) 53.0 (2.1)
10 91 91 92 93 93 92.1 (1.3) 95.0 (0.0) 91.0 (0.0) 94.0 (0.0) 95.0 (0.0) 87.9 (0.6)
15 96 92 94 96 96 95.2 (0.6) 95.0 (0.0) 92.0 (0.0) 94.0 (0.0) 95.0 (0.0) 87.3 (1.1)
C 5 - - - - - 98.7 (1.7) 96.8 (0.0) 96.8 (0.0) 96.8 (0.0) 96.8 (0.0) 56.5 (5.5)
10 - - - - - 100.0 (0.0) 100.0 (0.0) 100.0 (0.0) 100.0 (0.0) 100.0 (0.0) 83.5 (1.1)
15 - - - - - 100.0 (0.0) 100.0 (0.0) 100.0 (0.0) 100.0 (0.0) 100.0 (0.0) 92.3 (1.7)

Percentage of correctly predicted characters. The first column indicates the subject, the second column the amount of repetitions per character. The values in braces are the standard deviation. Subject A and B are datasets from BCI Competition III. Subject C originates from Competition II. We have omitted a direct comparison with related work on subject C because different methods are able to achieve 100% on all experiments.