Information Technology Reference
In-Depth Information
Table 8.1
Criteria to achieve identi
ability of sources in BSS methods based on AJD of SOS
Assumption
on the
sources
Covariance matrices (CM)
estimation
What is the energy
prole
Sufcient condition for identiability
Examples of data
Coloration
i. Lagged covariance
matrices,
ii. Fourier cospectral
matrices,
iii. CM estimated with a
lter bank
i. The source
autocorrelation
,
ii. The source
power
spectrum
,
iii. As in
ii.
The power spectrum of the source is
non-proportional to the power
spectrum of any other sources
Spontaneous oscillation with
characteristic power spectrum such as
posterior dominant rhythms (alpha),
somatosensory mu rhythms, frontal
midline theta, beta bursts, etc.
Non-
Stationary
CM estimated on
i. Different
time blocks
of
data
ii. Different
experimental
conditions
The variation of the
energy of the source
along the (i)
blocks or
(ii)
experimental
conditions
The variation of the source energy
along
(i)
blocks or
(ii)
experimental conditions
do not
correlate with the same variation of
any other sources
Blocks of data according to
physiological reactivity of EEG
oscillations (e.g., eyes close vs. eyes
open)
CM estimated before and after the event
in ERD/ERS
CM estimated on different peaks in ERP
(after averaging the ERP)
Active vs. control condition
-
