Information Technology Reference
In-Depth Information
of commercial and research grade (e.g. g.Tec MOBIlab) EEG hardware. The
outputs of these can be easily connected to sound and music programming envi-
ronments such as PD and Max through the use of UDP networking libraries like
open sound control (OSC) to create BCMI systems. However, no matter what
approach you take, it is crucial to understand the basic method and its limitations.
3.2.3 Detecting P300 ERPs Using the Averaging Method
Conventional methods for detecting ERPs represent some challenges for those
wishing to use them in musical contexts. This is because one of the core features of
ERPs, including the P300, is that when mixed with spontaneous EEG signals, or
what we might call background EEG, they are not very easy to spot. In most cases,
in order to detect ERPs we must remove the background noise that masks them.
The problem is that ERPs themselves look quite a lot like the spontaneous
potentials which make up the background noise, and occur in the same frequency
bands. Therefore, simply filtering the signals using standard methods will not work.
For these reasons, to remove the possibly random noise present in a time series of
EEG data, a series of stimuli are usually presented, and the responses are averaged.
This process reduces the amplitude of random signal components, and increases the
amplitude of similar signal components
the ERPs. To test this approach we can do
an oddball task, the basic process for which is detailed below (Fig. 3.2 ).
The Oddball task is a well-known P300 ERP paradigm (Polikoff et al. 1995 ). It
is useful for testing if an EEG system with tagged stimuli is able to elicit and detect
P300s. For the oddball task, two types of stimuli are required, for example, blue
circles and red squares. The stimuli are
flashed randomly on the screen at set
intervals. If we decide that the red square is to be the less common and therefore
more unexpected stimuli, the red squares are
flashed less often
for example at a
ratio of 10:2 for blue circles versus red squares. Each time a
flash is triggered, a
400 ms chunk of EEG data is stored and tagged to the stimulus. At the end of each
run, results for each stimulus type are averaged together. Both the blue circles and
the red squares will then have only one single averaged EEG chunk that represents
the average response of the brain for when there was either a blue circle, or a red
square. So in total, there are only two averaged chunks at the end of the oddball test.
If the averaged signal contains an amplitude peak between 200 and 600 ms after
the onset of the stimulus, and the average peak is greater than that of the averaged
peak in the other signal, it is judged to be a possible P300 target signal, as the target
would be the averaged signal with the highest amplitude. In the case of the oddball
task, if we know that the red square appears less often, we
d expect it to have a
higher average peak amplitude at that point. If this doesn ' it happen, it basically
means either the equipment failed, or the user blinked/moved too much, or was
thinking about something else. This sort of thing happens quite a lot, so we need to
'
 
Search WWH ::




Custom Search