Study on single-trial feature extraction method of single-channel visual evoked potential

Aiming at the low recognition rate problem in single-trial feature extraction of P300, a method based on orthogonal B-spline wavelet transform and Fisher linear Discriminant (FLD) was proposed. First, a method of coherent averaging combined with wavelet transform is introduced to preprocess the trial, and according to the time-frequency characteristics and time-locked relationship in VEP, a 8-dimension wavelet coefficient template representing P300 is constructed; then the features of the trial are extracted using the template; finally, a Fisher linear classifier is designed, which determines whether a single input is visual evoked EEG or not. Experiment results demonstrate that the method based on orthogonal B-spline wavelet transform and FLD has a good average recognition rate as high as 90% for the P300 identification.