Aiming at the prediction precision and applicability problem for the traditional software reliability prediction models, from the point of nonlinear time sequence, this paper presented a novel software reliability prediction model using RBF neural network based on empirical mode decomposition theory. In the paper, the fault data series obtained from software reliability test phase is decomposed into a series of intrinsic mode functions and a residue signal. Then a RBF network is constructed for an intrinsic mode function or the residual signal. Finally output of every prediction model is integrated into one output with equal weighted. Experimental results showed that the proposed model had higher precision of prediction and better applicability, compared with traditional software reliability models.