A tactile P300 brain-computer interface

Abstract

In this study, we investigated a Brain-Computer Interface (BCI) based on EEG responses to vibro-tactile stimuli around the waist. P300 BCIs based on tactile stimuli have the advantage of not taxing the visual or auditory system and of being potentially unnoticeable to other people. A tactile BCI could be especially suitable for patients whose vision or eye movements are impaired. In Experiment 1, we investigated its feasibility and the effect of the number of equally spaced tactors. Whereas a large number of tactors is expected to enhance the P300 amplitude since the target will be less frequent, it could also negatively affect the P300 since it will be difficult to identify the target when tactor density increases. Participants were asked to attend to the vibrations of a target tactor, embedded within a stream of distracters. The number of tactors was two, four or six. We demonstrated the feasibility of a tactile P300 BCI. We did not find a difference in SWLDA classification performance between the different numbers of tactors. In a second set of experiments we reduced the stimulus onset asynchrony (SOA) by shortening the on- and/or off-time of the tactors. The SOA for an optimum performance as measured in our experiments turned out to be close to conventional SOAs of visual P300 BCIs.

Keywords: BCI; BMI; P300; somatosensory; tactile.

Figure 1

Schematic overview of the tactor layout (top view) in the different conditions of Experiment 1. The tactors used are colored gray whereas the inactive tactors are white. In the FourPlus condition, only the gray tactors could be targets. The black tactors vibrated twice as often as the gray tactors.

Figure 2

Mean performance of the classification model in the four conditions of Experiment 1 as expressed by classification accuracy, or proportion of targets correct (A), classification accuracy corrected for chance (B) and bitrate (C). The stars in (A) indicate chance performance. Error bars represent standard errors of the mean.

Figure 3

EEG averaged across participants and presented separately for targets (solid lines) and distracters (dashed lines) in the four conditions of Experiment 1. Only data from electrode Pz is presented.

Figure 4

Mean performance of the classification model in the four conditions of Experiment 2 as expressed by classification accuracy (A), classification accuracy corrected for chance (B) and bitrate (C). The stars in (A) indicate chance performance. Error bars represent standard errors of the mean.

Figure 5

EEG averaged across participants and presented separately for targets (solid lines) and distracters (dashed lines) in the four conditions of Experiment 2. Only data from electrode Pz is presented.

Figure 6

Mean performance of the classification model in the four conditions of Experiment 2b as expressed by classification accuracy (A), classification accuracy corrected for chance (B) and bitrate (C). The stars in (A) indicate chance performance. LL: Long On–Long Off, LS: Long On–Short Off, MM: Medium On–Medium Off, MS: Medium On–Short Off. Error bars represent standard errors of the mean.

Figure 7

EEG averaged across participants and presented separately for targets (solid lines) and distracters (dashed lines) in the four conditions of Experiment 2b. LL: Long On–Long Off, LS: Long On–Short Off, MM: Medium On–Medium Off, MS: Medium On–Short Off. Only data from electrode Pz is presented.