Unsupervised speaker adaptation of DNN-HMM by selecting similar speakers for lecture transcription | IEEE Conference Publication | IEEE Xplore