Characteristics analysis of gearbox vibration response signals captured from vibrating acceleration sensor based on empirical mode decomposition (EMD) and statistical properties of nonlinear kurtosis is proposed. The vibration response signal is firstly decomposed into intrinsic mode function (IMF) by the empirical mode decomposition method. Then nonlinear Teager energy operator tracks the modulation energy of each IMF. The desired feature of statistical properties of vibration signals can be extracted from the coefficient nonlinear kurtosis value of intrinsic mode function. It is significant for the mechanical operation security to do some research on how to monitor operating state of gear and detect incipient faults as soon as possible. Experiment results have shown the feasibility and efficiency of the EMD and nonlinear kurtosis method in fault message diagnosis, and additionally, the algorithm is reliable to be implemented with fault detection.