Information Technology Reference
In-Depth Information
the name (AJDC). AJDC is simple, fast, and
flexible, allowing explicit modeling of
physiological and experimental a priori knowledge. We have argued that the suc-
cess of the source separation depends solely on
An appropriate choice of covariance matrices to form the diagonalization set;
￿
Their appropriate estimation.
￿
first requirement, we have provided guidance for the analysis of
continuously recorded EEG, event-related (de)synchronizations (ERS/D), and ERP.
While studies for the
To ful
ll the
first two cases are already available and well established, we
have here presented for the
first time the use of AJDC for ERD. We have conducted
a source analysis by means of BSS of the feedback ErrP in high cognitive load
conditions. In this experiment, we have used conditions that resemble those one can
find on real BCI experiments. Our results showed that the feedback-related potential
observed here shares the same characteristics as the FRN observed in gambling
tasks and the ERN observed in reaction time tasks. Indeed, all three error potentials
are notably characterized by a negative de
ection generated by the dorsal ACC, but
with different time of activation. A sharp analysis in the source space by means of
approximate diagonalization of covariance matrices has allowed the identi
cation
of three main components accounting for the differentiation between error and
correct trials. Two temporal (ERP) characteristics were identi
rst sharp
negativity (Ne) and a broad positivity (Pe). One frequential (ERS) characteristic
was identi
ed: a
ed as theta ERS at the same time that the Ne. This observation is in
accordance with previous
findings (Luu et al. 2004 ; Trujillo and Allen 2007 ) which
also pointed to the implication to oscillations in the theta band as an indicator of
response error-related potentials. Luu et al. (Luu et al. 2004 ) reported that the theta
band (4
7 Hz) is responsible for most variability of the ERN (57 %); meanwhile,
Trujillo and Allen ( 2007 ) reported a power increase in the theta band at a time
course similar to the Ne for erroneous responses. In this paper, we have observed
that the ErrPf is characterized by an important ERS in the theta band. This ERS
seems to occur at the same time as the negative evoked potential. This observation
leads to the question of the independence of these two components. Indeed, even if
they occur simultaneously, they may represent different manifestation of the same
neuronal process. Blind source separation coupled with source localization (sLO-
RETA) has allowed the identi
-
cation of two spatially distinct sources, one
accounting for the temporal component (BA24) and the other for the frequency
component (BA6). Statistical analysis at source level validated this separation with
a signi
cant temporal activity only for the
first source exhibiting a signi
cant ERP
at the time of Ne and Pe and a signi
cant ERS only for the second source in the
theta band. The fact that these two sources are uncorrelated and spatially segregated
suggests that these two phenomena do not re
ect the same neuronal process. This
point is of great interest for BCI applications and for the online detection of the
ErrPf since they may therefore provide independent information for classi
cation.
In fact, up to now in BCI, only the negative wave (Ne) has been used as a feature
for classifying the ErrPf. Our results suggest that one could use both the ERP
 
Search WWH ::




Custom Search