Abstract
An experiment was conducted to determine the feasibility of using waveform features of the single trial transient evoked cerebral potential as a predictor of a simple “YES or NO” decision response. Six subjects, who were tested in a stimulus identification test, were shown a 100-trial sequence of instruction stimulus sets to which they decided by a button push whether the stimulus agreed with the instruction. The Electroencephalogram (EEG) was recorded during the test, along with a stimulus trigger pulse used to align the EEG with the stimulus. The coefficents of a Linear Predictor Coding model fit to the post-stimulus, evoked potential signals were used as inputs for training and testing a supervised, hidden layer, neural network representation of the decisionmaking process. The resultant error rate computed using the “leaving one out” procedure, varies from 32-% to 46-% for predicting the decisions made by the subjects.
Get full access to this article
View all access options for this article.