Fig. 1From: A Bayesian view on acoustic model-based techniques for robust speech recognitionBayesian network representation of a a conventional HMM and b an HMM incorporating latent feature vectors. Subfigure b is based on [20]Back to article page