Information Technology Reference
In-Depth Information
9.5.2 H ¨ lder Exponent Criterion and Alpha Criterion
Previously, other approaches to obtain a summarized data matrix in two dimensions
have been tested on similar signals (V
zard
2010
). The goal was to obtain an
approach which allows separating the two alertness states and reducing the inter-
individual variability observed. One of these approaches was based on the use of
the H
é
lder exponent, (Jaffard and Meyer
1996
,
Levy Vehel and Seuret
2004
), is a tool to measure the regularity of a signal at a
given point. The smaller the H
ö
lder regularity of the signal. The H
ö
lder exponent (respectively, large) is, the more
irregular (respectively, smooth) is the signal. The H
ö
ö
lder exponent was estimated as
de
ned in Legrand (
2004
). The aim was to summarize the signal recorded by an
electrode in its global regularity. An average of H
lder exponents for each point of
the signal provided by an electrode was calculated.
Another approach was to analyze the alpha wave content in signals. Alpha
rhythm is the classical EEG correlate for a state of relaxed wakefulness. When the
person is relaxed, the neurons are synchronized and operate at a particular and
identical rhythm. This rhythm appears to be responsible for the more pronounced
appearance of Alpha waves (Niedermeyer and Lopes da Silva
2005
). When the
person is forced to perform a task that can break the relaxed state, the functioning of
neurons vary widely. They seem to act by groups which do not work at a similar
rhythm. Alpha waves are then masked by the more pronounced appearance of other
waves (such as Beta waves). Thus, the idea was to measure the proportion of alpha
waves in the signal (alpha waves divided by the sum of all waves: alpha, beta, theta,
and delta).
These two approaches gave a data matrix in two dimensions like that obtained
with the slope criterion. However, they did not seem to work as well as the matrix
of slopes to discriminate the two states of vigilance (V
ö
é
zard
2010
). Therefore, the
slope criterion is investigated in this topic chapter.
9.5.3 Preliminary Results
The relevance of the slope criterion is illustrated in Figs.
9.15
and
9.16
. Figure
9.15
provides for each participant, in his state of
“
normal
”
alertness and his state of
“
alertness, the sum of the slope criterions on all electrodes. It appears that
for a given individual, the slope criterion is almost always lower when the indi-
vidual is in the normal state than when he is in the relaxed state. Thus, by com-
paring, for a given individual, the values of the slope criterion for the normal and
relaxed states it is possible to effectively distinguish the two states. However, for a
new individual, a single record is known and the problem remains unsolved.
Figure
9.16
shows for each electrode the sum of the slopes of the participants in a
“
relaxed
”
state. The previous obser-
vation is also true at the electrode level. In fact, for a given electrode, the slope
criterion is higher when considering the record obtained by this electrode after the
normal
”
alertness state and participants in a
“
relaxed
”
