Temporal principal component analysis (t-PCA) has been widely used to extract event-related potentials (ERPs) at the group level of multiple subjects' ERP data. The key assumption of group t-PCA analysis is that desired ERPs of all subjects share the same waveforms (i.e., temporal components), whereas waveforms of different subjects' ERPs can be variant in phases, peak latencies and so on, to some extent. Additionally, several PCA-extracted components coming from the same ERP dataset failed to

more »

... e statistically analysed simultaneously because their polarities and amplitudes were indeterminate. To fill these gaps, a novel technique was proposed and employed to extract desired ERP from single-trial EEG dataset of an individual subject. Firstly, the dataset of all trials and all conditions of one subject were stacked across electrodes to form a matrix. Secondly, the temporal and spatial PCA-components were extracted from single-trial EEG dataset matrix for each subject by t-PCA and Promax rotation. Thirdly, the desired components were selected and projected to the electrode fields simultaneously to correct their variance and polarity indeterminacies. Next, single-trial EEG datasets of the back-projection were averaged to generate the waveforms of desired ERP for each subject and then amplitudes of the desired ERP were quantified. The yields indicated that the proposed approach can efficient exploit the temporal and spatial information of single-trial EEG data and can temporally filter the data to extract the desired ERPs for an individual subject.