Neural network classification of spatio-temporal EEG readiness potentials

Abstract

The detection of spatio-temporal scalp EEG patterns associated with voluntary motion preparation towards the development of a brain-computer interface (BCI) is explored. The rationale for the use of a spatio-temporal approach to this detection problem is explained. The need for a temporal or dynamic classifier is confirmed by demonstration of the lack of robustness in static neural network classifiers with respect to time alignment of the patterns under analysis. The results from dynamic classifiers, such as the Time Delay Neural Network (TDNN) and the Gamma Neural Network are presented in terms of their Receiver Operating Characteristic (ROC) Curves.