Abstract

Classification of stellar spectra by human experts, in the past, has been subjective, leading to many non-unique databases. However, with the availability of large spectral databases, automated classification schemes offer an alternative to visual classification. Here, we present two schemes for automated classification of stellar spectra, namely, chi-square – minimization and Artificial Neural Network. These techniques have been applied to classify a complete set of 158 test spectra into 55 spectral types of a reference library. Using these methods, we have successfully classified the test library on the basis of a reference library to an accuracy of two spectral subclasses. Such automated schemes would in the future provide fast, uniform, and almost on-line classification of stellar spectra