Error-related potentials (ErrP) are changes in EEG signals that can be elicited when users perceive errors. Recognising ErrP can help improve the performance of P300 Speller. However, it is a hard task due to the low signal-to-noise ratio and high variability from trial to trial. In this study, a novel method is proposed to reveal the difference in time series pattern between single trial EEG epochs that contain ErrP and ones that do not contain ErrP. For testing our approach, four healthy subjects were recruited to take part in P300 Speller experiments with feedback. The performance was evaluated with sensitivity and specificity metric. The attained results were 54.7%, 89.6% respectively. The comparison to other methods showed its effectivity.