[A method for auto-recognizing the stars based on spectral feature].

The LAMOST project, the world's largest sky survey project being implemented in China, urgently needs an automatic stars recognition and classification system. This paper presents a method for auto-recognizing the stars based on spectral feature. This method consists of three main steps: First, the integral information of spectral lines is calculated and the stellar Balmer lines are detected by using the wavelet features of spectral lines. Then, the characteristic frequency of M-type stars and the locations of absorption bands are obtained accurately through the wavelet features of absorption bands. Finally, based on the results of the former step, the emission-line stars, M-type stars and early-type stars can be recognized. The extensive experiments with real observed spectra from the SDSS DR4 show that the method can robustly recognize stellar spectra, the correct rate of the emission-line stars is as high as 97.5%, the correct rate of M-type stars is as high as 98.1% and the correct rate of early-type stars is as high as 96.8%. The error rate of the quasars and the galaxies is less than 2%. This method is designed to automatically recognize stellar spectra with relative flux and low signal-to-noise ratio, which is applicable to the LAMOST data.