Neural network classification of EEG using chaotic preprocessing and phase space reconstruction

A cognitive mode mapping system is developed that analyzes and classifies electroencephalograph (EEG) signals recorded from four sites of a subject´s brain. The subjects produce this EEG data while performing five selected cognitive tasks. The objective of the system is to identify these tasks based on the salient features embedded in the raw EEG signals. Also, due to the demanding requirements of some environments (such as jet fighter cockpits), achieving the state recognition in near real-time is critical. Initial experiments show the system is able to correctly classify the EEG signals from the subjects 100% of the time. The classification delay is approximately 15 seconds due to the initial 10 seconds of data gathering and 5 seconds of network feedforward processing delay. It is also found that the trained network can recognize the subjects´ EEG days after the initial training took place