Povey, DanielBurget, LukasAgarwal, MohitAkyazi, PinarFeng, KaiGhoshal, ArnabGlembek, OndrejGoel, Nagendra KumarKarafiat, MartinRastrow, AriyaRose, Richard C.Schwarz, PetrThomas, Samuel2014-11-192014-11-192014-11-19201010.1109/ICASSP.2010.5495662https://infoscience.epfl.ch/handle/20.500.14299/108950We describe an acoustic modeling approach in which all phonetic states share a common Gaussian Mixture Model structure, and the means and mixture weights vary in a subspace of the total parameter space. We call this a Subspace Gaussian Mixture Model (SGMM). Globally shared parameters define the subspace. This style of acoustic model allows for a much more compact representation and gives better results than a conventional modeling approach, particularly with smaller amounts of training data.Speech RecognitionHidden Markov ModelsGaussian Mixture Modelstext::conference output::conference proceedings::conference paper