Rapid feature space MLLR speaker adaptation with bilinear models | IEEE Conference Publication | IEEE Xplore