In view of the serious distortion of the low frequency trend term signal in the original vibration acceleration signal when it is transformed by mathematical integration， a method combining least squares fit （LSF） and empirical mode decomposition （EMD） is proposed to filter the interference signal in the original signal. In this method， the intrinsic mode function （IMF） obtained from empirical mode decomposition is reconstructed after removing the trend term to achieve the purpose of signal noise reduction. The method is used to reduce the noise of the analog signal and the vibration data of a certain type of dry vacuum pump， and then the signal integral transformation is carried out. The comparison shows that this method can make up for the deficiency of single method in processing low frequency trend term of signal. It improves the reliability of vibration signal analysis and provides a good foundation for further research on fault monitoring and diagnosis of dry vacuum pump.