| |
Abstract:
Due to a variety of recent advancements, particularly in our
understanding of the EEG patterns associated with voluntary
movement and development of artificial pattern recognition
techniques, it is now possible to construct BCI systems which can
recognize meaningful information about a user's desires based only
on EEG data. BCIs have already been built which can allow
individuals to move a cursor by changing (for example) their mu
rhythm, a free running EEG associated with movement (Wolpaw et al.,
1994; McFarland 1998). These systems have tremendous potential for
allowing severely disabled individuals to communicate, and more
advanced BCIs could be useful in a wide variety of applications.
This project drew on two concurrent research projects exploring
changes in the human mu rhythm (Pineda et al., 1998) and readiness
potential (Hughes et al., 1998). New findings demonstrating changes
in EEG activity preceding voluntary movement of either one limb or
combinations of limbs will first be presented. In particular, the
readiness potential preceding combinations of limb movements is
significantly larger for difficult movements. We also present
results comparing the effectiveness of several pattern recognition
approaches (such as neural networks, Hidded Markov Models, and
independent component analysis) and discuss their
performance.
|
