PROJECT OBJECTIVES

• Implement several standard speech recognition front-ends:
  • Mel-scaled cepstra,
  • Perceptual linear prediction,
  • Filter bank amplitudes,
  • Delta and delta-delta features.

• Interface the front-end speech signal processor into a real LVCSR system.

• Develop a graphical user interface to be used as an educational tool in a speech recognition course.

• Evaluate the effectiveness of each algorithm on a real speech corpus.