[Univ of Cambridge] [Dept of Engineering]

Speech Practical

This practical is concerned with phoneme-based continuous speech recognition using Hidden Markov Models (HMMs). The aim is to train a set of speaker independent models on a subset of the TIMIT Acoustic-Phonetic speech database and then measure the performance on a second different subset. The software required to train the models is provided but the recogniser is to be written by you in C++.

The speech practical handout from is available in [ps] [pdf]

Any queries, problems, or errors in the handout, please contact me by email mjfg@eng.cam.ac.uk

Important dates are:

  • End Week 4 Lent Term: first part of practical checked by a demonstrator.
  • 22nd March 2005: hand in speech practical report.
top

[ Cambridge University | CUED | MIL | Home]