Information Technology Reference
In-Depth Information
Fig. 3.2
The oddball paradigm. The approach represented
above
is slightly more ef
cient than the
standard approach as instead of simply calculating which averaged chunk has the highest
amplitude, here we calculate the area underneath the curve containing the highest positive
amplitude peak (for details see Fig.
3.1
), which is more robust as it helps to reduce inaccuracies
introduced by high frequency amplitude spikes and other noise/unwanted signals
have ways to account for it, such as discounting epochs of data with extraordinarily
large peaks consistent with head movement and blinking.
Of course, if we want to be able to choose between three or more different
possible choices, we can have a grid containing all possible choices, and
ash each
choice randomly an equal number of times, whilst we ask the user to stare at their
chosen target. Success rates for this method vary wildly depending on a range of
factors, including the skill of the user, how much they are concentrating, the quality
and/or position of the electrodes, the ease with which the stimuli can be seen, the
number of times each stimulus is
flashed, which algorithms are used for detection,
how the signal is analysed prior to averaging, the accuracy of the signal baseline,
and any number of less important but still occasionally signi
cant variables.
However, almost all of these variables also apply to all other forms of EEG BCI.
Despite these disadvantages, ERP detection is a surprisingly accurate method for
BCI, with detection rates sometimes reaching over 90 %, meaning that 90 % of the
time, the system is able to detect which stimuli is triggering the P300. As the
method favours actual conscious attention measured with reference to the time
when stimuli
flashed, it is one of the best approaches for EEG-based BCMI in terms
of accuracy and usability. Later in this chapter we will discuss a number of possible
applications for using P300 BCI in music contexts. In addition we will also detail
ways we can signi
cantly improve both the speed and usability of P300 ERPs
for BCMIs.
