Noise in dynamical systems is usually considered a nuisance. But in certain nonlinear systems, the presence of noise can in fact enhance the detection of weak signal. This phenomenon, called stochastic resonance (SR), may find useful application in physical, technological and biomedical fields. A novel approach to detecting weak periodic signal using stochastic resonance theory is presented. This method is analyzed and validated by simulated signal. The result shows that this method is simple, robust and reliable. The weak sinusoid signal of lower signal-to-noise ratio can be reliably extracted from loud noise. The detection approach of weak signal based on SR indicates a promising prospect for mechanical fault diagnosis.