2005AJ....129..669L

An empirical method of modeling the stellar spectrum of galaxies is proposed, based on two successive applications of principal component analysis (PCA). PCA is first applied to the newly available stellar library STELIB, supplemented by the J, H, and K_smagnitudes taken mainly from the Two Micron All Sky Survey. Next, the resulting eigenspectra are used to fit the observed spectra of a sample of 1016 galaxies selected from the Sloan Digital Sky Survey Data Release 1 (SDSS DR1). PCA is again applied to the fitted spectra to construct the eigenspectra of galaxies with zero velocity dispersion. The first nine galactic eigenspectra so obtained are then used to model the stellar spectrum of the galaxies in SDSS DR1 and synchronously to estimate the stellar velocity dispersion, the spectral type, the near-infrared spectral energy distribution, and the average reddening. Extensive tests show that the spectra of different types of galaxies can be modeled quite accurately using these eigenspectra. The method can yield stellar velocity dispersion with accuracies better than 10% for the spectra of typical signal-to-noise ratios in SDSS DR1.