Human'S Performance Prediction with Wavelet Analysis of EEG Data

Predicting a person's performance in an oddball experiment using their EEG data is particularly challenging as only the part of the EEG signal preceding the reaction time can be used. We consider here the use of wavelet coefficients as features for discriminating between a person's quick and late reaction. Our motivation is the relation between the brain rhythms (certain frequency bands of the EEG signals) and a person's ongoing cognitive processes. We demonstrate that wavelet coefficients of specific EEG channels, scales and time shifts have an enhanced discriminating capability and present an algorithm for predicting a person's quick or late reaction in an oddball experiment. The results are compared with those acquired with equivalent time domain features and are found superior for a number of cases.