STK Toolkit is a set of tools for training speech recognizers based on Hidden Markov models and for the recognition itself. It is compatible with the standard HTK Toolkit from University of Cambridge, but brings some substantial enhancements, especially in the discriminative training of acoustic models. Tools for keyword spotting are also part of STK. The toolkit is being developed thanks to support of EU-sponsored project AMI.
New trends in research and application of voice technology (GA102/05/0278)