Multi-frame rate based multiple-model training for robust speaker identification of disguised voice | IEEE Conference Publication | IEEE Xplore