Based on the in-depth study on morphological wavelet with permutation entropy，a new method for gear of transmission fault recognition is proposed. Firstly，the definition of morphological wavelet is introduced，and the morphological Haar wavelet is used to pre-process the measured gear of transmission vibration signal. Then，the permutation entropy is used as the eigenvalue of gear fault to extract the vibration signal, which included four working conditions: normal，slight-worn, medium-worn and broken teeth. Finally，according to different faults corresponding to different permutation entropy distributions，the various fault states are classified，and the permutation entropy distributions of non-denoised signals are compared. The example of gear fault recognition proved that the combination of morphological wavelet and permutation entropy could effectively improve the ability of gear of transmission fault classification.