On the jointly unsupervised feature vector normalization and acoustic model compensation for robust speech recognition | Publicación