This paper deals with the numerical simulation of stochastic resonance (SR) for detection weak signal. SR is a nonlinear phenomenon generally occurring in dynamical bistable systems excited by noisy periodic signals. Through SR, weak signals immerged in heavy noise can be amplified and optimized by assistance of noise, due to the interplay between noise sources and nonlinearity of the system. We present the numerical method to solve the bistable SR model based on Runge-Kutta algorithm, and propose a new kind of approach for detecting weak signal, which can effectively detect both period and aperiodic weak signals corrupted by the strong noise background.