| Recent advances in computer hardware and signal processing have made it feasible to use human electroencephalograph (EEG) signals to communicate with a computer. Locked-in patients now have a new means(that is Brain-Computer Interface technique) to communicate with the outside world. But Brain-Computer Interface (BCI) systems still suffer from the slow communication rate on the order of 5-25 bits/minute. The signal processing of EEG is one of the important factors to decide the communication rate of such systems. The P300 evoked potential is a positive wave in the EEG signal peaking at around 300 milliseconds after task-relevant stimuli and it can be used as a binary control signal for the BCI system. But due to the non-stationary feature and the low amplitude of evoked potentials (EPs) in comparison with the ongoing EEG, and the frequency band of EPs overlapped with the ongoing EEG, so it is hard to extract EPs from the background EEG with conventional signal processing method. Due to the good localization feature of the wavelet transform both in time plane and in frequency plane, we present a wavelet-based denoising method for extracting EPs in single trails. Futhermore, the selection criterion of the trails containing the P300 wave according to the occurrence of a positive delta (0-4Hz) coefficient after stimulus onset between 400 and 600 ms was presented in this paper. The results show that we can distinguish the evoked responses P100-N200 and P300 in most of the trails. We also calculated the cross-correlations between the denoised single trails and the average denoised trails for each subject, and finding that 85% trials whose cross-correlation with the average was larger than 0.45. The results show that only 74% of the target sweeps contain P300 component according to this criterion. These results are hard to achieve with conventional Fourier methods.A perfect PC-based EP wavelet analysis system was designed in this paper.
|
PDF Full Text Request