The Independent Component Analysis (ICA) method has a good application in EEG. But ICA algorithm is mostly a combination of static model and batch optimization algorithm. Considering the non-stationary characteristics of event-related potentials, the Dynamic Independent Component Analysis (DICA) based on kurtosis maximization was proposed to single-trail extract the event-related potential N400 which is induced by the Chinese idioms whose last word is incorrect. The DICA method achieved satisfactory results.