Combining EigenVoices and structural MLLR for speaker adaptation | IEEE Conference Publication | IEEE Xplore