Background: The α particle has a strong ionization but weak penetration ability

and some factors

such as α radioactive source self-absorption

incomplete charge collection

electronic noise and statistical fluctuation

etc. have significant influence on the experimental α spectra synthetically

which can lead to a long tail at low-energy side and serious spectra noise. Purpose: The aim is to reduce the statistical fluctuation and electronics noise in detector

so as to preprocess data for the deconvolution of α spectra. Methods: In this study a portable α spectrometer with a Au-Si surface barrier detector was applied to obtain α spectra of 239Pu under vacuum conditions of ?0.01 MPa

?0.02 MPa

?0.03 MPa

?0.04 MPa and ?0.05 MPa. A discrete wavelet transform(DWT) method for α spectra smoothing and a new smoothing evaluation method were proposed

and tested on those spectra data for comparative study. Results: The results showed that DWT method had less damage on the peak compared with least squares polynomial smoothing method. The D(r) and ξ values of DWT were greater and χ2 values were closer to 1 compared with those of least squares polynomial smoothing method. Conclusion: The DWT method is better than the traditional least squares polynomial smoothing method in smoothing low-level and high noise α spectra.