To reduce noise in mechanical vibration signal, a new adapted denoising method is proposed in paper that is the joint of ensemble empirical mode decomposition (EEMD) and least mean square algorithm (LMS). First, the algorithm performance of the LMS with fixed step and fixed filter order, variable step size and variable order were studied, it is proposed that convergence direction of algorithm is up to the expectation of least mean square error because of order or step changing in iterative process, a kind of jointing variable order and variable step LMS (VSVT-LMS) was developed, and Using EEMD, the original signal was decomposed, so that every mode component would be narrowband, and then through VSVT-LMS denoising, the algorithm instability of the LMS for wideband signals, but also the EMD threshold selection problem were avoided effectively. Simulation and vehicle actual signal were analyzed and it shows that the new method has a good adaptive characteristics and accuracy, and the feasibility of the method was verified in engineering.