A Stratified Approach for Automatic Stellar Spectra Classification

In this paper we present an automatic stellar spectra classification system to achieve two goals: one is the main-class and subclass classification, the other is luminosity type recognition. The system is composed of 3 functional units, namely main-class and subclass classification, continuum spectra normalization, luminosity type determination. This work´s main novelties are 2 folds: 1) A stratified approach is proposed for classification. The main-class and subclass of star is firstly determined, and then according to the main-class type, luminosity type is recognized through luminosity type determination unit. The luminosity type determination unit is composed by seven models based on PLS method each corresponding one of the seven mainclass type. 2) We use the non-parameter regression method to robustly determine the main-class and subclass, and PLS regression method to determine the luminosity type. Experiments show that our system is feasible for the low-resolution spectra classification.