Abstract: Independent component analysis (ICA) is a novel blind source separating method, which can extract individual signals from mixtures of signals. According to the peculiarities of electroencephalogram (EEG) and event related potentials (ERP), a fast algorithm based on separating characters of ICA is used to extract the useful information of ERP from EEG. A brief illustration of the basic principles and common criteria of ICA is first given, together with explanations of the ICA algorithm. ICA is then applied in the separation of simulated data set and extracting ERP from EEG. Satisfactory results obtained in both aspects illustrate the performance and validity of the algorithm.