Daniel Povey

An indefatigable enthusiast of speech recognition technology, Povey has to his credit pioneering work on “discriminative training” of Hidden Markov Models for speech recognition, having developed many of the standard techniques.

Dan Povey was appointed Associate Research Scientist at the Center for Language and Speech Processing at the Johns Hopkins University in 2012. Povey researched speech recognition at Microsoft and IBM between 2003 and 2012. An indefatigable enthusiast of speech recognition technology, Povey has to his credit pioneering work on “discriminative training” of Hidden Markov Models for speech recognition, having developed many of the standard techniques. He is currently occupied with finding ways to represent generative models more compactly (such as the Subspace Gaussian Mixture Models approach), for purposes of more robust parameter estimation. Developing the open-source Kaldi speech recognition toolkit is another Povey passion, which has given him a thorough understanding and practice of speech recognition technology. Povey earned a doctorate from Cambridge University in 2003. He also acquired a BA in Natural Sciences Tripos from Cambridge.