The signal-to-noise ratio of vibration signal of micromotor is low,and the environmental noise is complex.The key step of quality inspection is to remove the noise signal effectively.Aiming at the problems of poor continuity and unsatisfactory denoising effect of traditional wavelet de-noising threshold function,an improved wavelet threshold function based on sample entropy was proposed,which could automatically adjust the threshold function according to the degree of signal confusion.The simulation results show that in the low SNR environment,the improved threshold function based on sample entropy is better than the traditional threshold function and the ordinary improved threshold function,and the signal-to-noise ratio is significantly improved.The noise reduction and feature extraction of the abnormal sound signal of the micromotor were carried out,and the training test was carried out combined with the SVM classifier.The experimental results show that the improved wavelet de-noising algorithm can be used to effectively remove the environmental noise of the motor signal,to extract the effective signal features,so the performance of the motor can be accurately judged.This method will provide theoretical basis and support for large-scale quality inspection of micromotor manufacturers.