EEG signals contain rich information on brain activity, reflect the electrical activity of brain tissue and brain functional state. By studying the EEG, which help a lot for disease diagnosis and clinical research. Yet the popular algorithm for EEG analysis requires further improvement in effect. Regarding this problem, this paper introduced a improved Independent Components Analysis(ICA) to EEG processing, and a novel method is proposed. The step and feature of the algorithm proposed is described in detail. Finally the simulation experiment demonstrates the effectiveness of the algorithm.