In order to eliminate the noise signal generated by the vibration signal of the hobbing spindle in the machining process due to the environmental influence,a signal noise reduction method based on Savitzky Golay-WPT was proposed. The original signal was calculated to obtain the optimal wavelet packet decomposition tree; according to the optimal decomposition tree,the wavelet packet transform was carried out to obtain the wavelet packet coefficients; the threshold function was used to filter the wavelet packet coefficients; the least square fitting method was used to reconstruct the coefficients after wavelet packet screening. The results show that compared with the traditional wavelet packet and CEEMDAN,the performance of the proposed method is improved by 31.35% and 22.71% respectively; in the actual processing data,compared with the traditional wavelet packet method,by using this method,the peripheral interference of the center frequency can be reduced,and the central frequency features is more prominent and the noise reduction effect is more obvious.