A two-level approach for speaker recognition using speaker-specific-text | IEEE Conference Publication | IEEE Xplore