Information Technology Reference
In-Depth Information
Fig. 3.4: (
a
) The simulated channel 2 consists of a time-varying (among different
participants) theta linearly modulated signal (length 2 s) occurring at a fixed latency
and a gamma linearly modulated signal (length 3 s) mixed with quasi-white noise.
Quasi-white noise also covers the interval between the modulated signals. (
b
)The
wavelet PS time-frequency representation picture. The significant regions over the
time-frequency transform are indicated by the contours. The significant signal seg-
ments (contours) are successfully discriminated from the white noise background.
on the significant regions and is able to detect and correctly account for the energy
content of the selected regions.
3.4 Discussion
Using the wavelet transform method on EEG signals, cortical activation evaluation
is normally performed by means of comparing a target task (while participant is
engaged with a difficult cognitive task or reflects certain pathology) and compares
it with a rest condition. This method, in contrast to traditional spectral ones, can
estimate changes between EEG signals without being bounded to the stationarity
assumption and can provide information for the entire time evolution of the signal.
The simulation test and the results presented justify the suggestion that relevant
characteristics are temporally localized in the most significant regions (contours in
the WT scalogram), rather than in the entire segment length of the EEGs. The
Global
PS only partially encapsulates the significant information, since there is significant
frequency leakage between the bands due to the transient response of the time-
frequency filter in different frequencies. Using such features, both channels 2 and 5
in the simulated case induce activity in almost every band. However, the proposed
