Toward speaker-independent recognition-by-synthesis | IEEE Conference Publication | IEEE Xplore