Speaker recognition with region-constrained MLLR transforms | IEEE Conference Publication | IEEE Xplore