105. ANALYSIS OF EEG DURING PLANNING AND EXECUTION OF NATURAL GRASPING TASKS

Department: Electrical & Computer Engineering
Faculty Advisor(s): Virginia De Sa

Primary Student
Name: Paul Sterling Hammon
Email: phammon@ucsd.edu
Phone: 858-822-2421
Grad Year: 2008

Abstract
A brain-computer Interface (BCI) is a system which translates brain states into computer-controlled actions. A complete BCI system includes: user mental task, brain signal acquisition, signal processing and feature extraction, brain state classification, user feedback, and performance of a computer-controlled action.

Our goal is to improve BCI systems by jointly developing the mental task and feedback components of the system to simultaneously improve ease of use and enhance the strength of associated brain signals. We previously presented a first step towards this goal by decoding reaching target from electroencephalographic (EEG) signals during the planning and executing of natural movements (Hammon et al., 2008).

In this work, we expand upon our previous efforts by including grasping of natural objects. We describe a delayed grasping task in which subjects interact with an object in one of three ways: touch, precision grip, and power grip. Our task involves two identical objects located to the left and right of the subject. Trials are randomized by target location and grasp type, allowing us to compare task EEG based on location in space (left, right) and type of object interaction (touch, precision grip, power grip).

We simultaneously record 64-channel EEG and 3D locations of the fingers, hand, and arm. EEG data are processed to remove motion artifacts and then analyzed during the delay and movement periods of the trial. Signal processing and classification involves several steps: feature extraction, dimensionality reduction, training of individual classifiers, and then combining classifiers to create meta-classifiers. With this approach we are able to build classifiers to compare BCI classification performance for reaching and grasping tasks both during planning and performance of actual reaching tasks. Our results will guide development of an on-line BCI system based on the imagination of natural movements.

« Back to Posters or Search Results