Abstract: Based on the learning mechanism of OLAM network and the theory of γspectrum analysis, an artificial neural network to carry out the activity analysis of nuclides was constructed. For getting γspectrum of the nuclide of a simple substance to train the net, the characteristic spectra of the nuclide were simulated with the scale equations given in the spectrum analysis software. The welltrained net was used to calculate the activities of nuclides 238U, 226Ra, 232Th, 40K and 137Cs in the samples of environmental soil. By comparing the results with the SPASG calculation, the average relative deviation is 2.24%, and this indicates that both the net construction and the characteristic spectrum simulation are successful.