A new method with Radial Basis Function Neural Network based on the EMG-Model ( EMG-RBFNN) was proposed for resolution of the overlapping chromatographic peaks . The genetic algorithm ( GA) was applied to train EMG-RBFNN so that it had the ability to learn its structure. Although GA has the advantage of robustness and globalism , it may stop a local optimization if population is less. But EMG-Model has low efficiency, time of resolution must be long if population is large. For increasing the efficiency of the EMG-Model , this paper designed a algorithm : first use RBFNN , and then use EMG-RBFNN.