Information Technology Reference
In-Depth Information
8.7
A Study on Error-Related Potentials
We now turn to the illustration of the AJDC method by means of a new study on
error-related potentials (ErrP). We show that BSS analysis increases the speci
city
and sensitivity that can be obtained working at the sensor level, increasing as a
consequence the single-trial classi
cation rate. ErrPs are a family of event-related
potential (ERP) that can be elicited after the commission of an error,
firstly reported
in Miltner et al. (
1997
) as associated to receiving external negative feedback after
error commission. This feedback error-related potential (ErrPf) is characterized by a
negative de
ection peaking between 250 and 400 ms with a fronto-central scalp
distribution. The authors named it the feedback-related negativity (FRN) and put it
in relation to the response error-related negativity (ERN) that had been previously
reported (Felkenstein et al.
1991
; Gehring et al.
1993
), also characterized by a
negative de
ection. Initially, the ErrPf has been studied prevalently in the case of
gambling tasks with monetary gain and loss. More recently, it has attracted much
attention in the brain
computer interface (BCI) community because its online
detection provides a unique opportunity to automatically correct erroneous BCI
operations, effectively increasing the consistency and transfer rate of a BCI system
(Farquhar and Hill
2013
). In order to do so, accurate online single-trial ErrP
detection is necessary. Here, we contribute along this direction in two ways: (1) We
design a new experimental protocol in order to study single-trial ErrPf detection in a
controlled situation that mimics actual BCI operation and (2) we apply the AJDC
source analysis in order to better characterize this potential, hence increasing the
accuracy of its online single-trial detection.
-
(1)
New Experimental Protocol
In all previous studies on single-trial detection of ErrP for integration of a control
loop in a BCI system, the involvement of the participants is very far from the
involvement of participants during BCI operation, that is, as such, they lack eco-
logical validity. In particular, in previous studies, the feedback is the main focus of
the subject, while in actual BCI operations, receiving such a feedback is only a
small part of a complex cognitive task. Furthermore, previous studies have mainly
returned shame feedback, that is, feedback completely unrelated to the performance
of the subject. Finally, the subject-speci
c control capability of a BCI system has
not been taken into consideration. Here, we study the feedback-related potential in
the case of a memory task, with no monetary gain or loss. The feedback is returned
when the subject gives the answer, and no reward is given to the subject except a
score; thus, our participants have no other interest besides their own performance.
Such an experimental protocol allows us to study the ErrPf in a real
“
error versus
correct
condition. The protocol we use is a memory task inducing a high cognitive
load. The subject is continuously engaged in a demanding task (and not only on the
feedback presentation), mimicking the actual conditions of a BCI use, where focus,
concentration, and attention are essential requisite for successful BCI operation.
”
