The rapid development of astronomical observation has led to many largesky surveys such as SDSS (Sloan digital sky survey) and LAMOST (large sky areamulti-object spectroscopic telescope). Since these surveys have produced verylarge numbers of spectra, automated spectral analysis becomes desirable andnecessary. The present paper studies the spectral classification method based onBayes decision theory, which divides spectra into three types: star, galaxy andquasar. Firstly, principal component analysis (PCA) is used in featureextraction, and spectra are projected into the 3D PCA feature space; secondly,the class conditional probability density functions are estimated using the non-parametric density estimation technique, Parzen window approach; finally, theminimum error Bayes decision rule is used for classification. In Parzen windowapproach, the kernel width affects the density estimation, and then affects theclassification effect. Extensive experiments have been performed to analyze therelationship between the kernel widths and the correct classification rates. Theauthors found that the correct rate increases with the kernel width being closeto some threshold, while it decreases with the kernel width being less than thisthreshold.(201008282)资助