Hard-mask missing feature theory for robust speaker recognition | IEEE Journals & Magazine | IEEE Xplore