The term spectral feature is a catch-all for any recognizable feature of a spectrum, ideally providing evidence regarding the characteristics of the source. It might refer to an emission line or absorption line, or the line shape, or a spectral band, or a Balmer jump, or a vegetation red edge. Some lines are common in galaxies and recognizable even when redshifted, offering a means to estimate distance using the Hubble constant.
The term spectral feature extraction is used in discussions of computer techniques aimed at analyzing spectra, allowing and population groupings and examples of specific phenomena to be selected from extensive survey data. Machine learning methods are of interest.