Section 5 : Speech Recognition as an application of HMM

Commentary

Section Goals

• To introduce techniques related to speech recognition, such as the probabilistic acoustic model, language model, and speech recognizer.

Learning Objectives

Learning Objective 1

• Describe the different components of speech recognition tasks.
• Explain the principles of probabilistic models in speech recognition and speech recognizers.
• Draw a system structure diagram to illustrate the information flow and process components for speech recognition.
• Explain the following concepts or terms:
  • Speech recognition
  • Speech understanding
  • Acoustic model
  • Language model
  • Bigram and trigram model
  • Pronunciation model
  • Phone model
  • Coarticulation
  • Continuous speech
  • Segmentation
  • Speech recognizer

Objective Readings

Required readings:

Reading topics:

Speech Recognition (see Section 23.5 of AIMA3ed)

Rabiner, L. R. (1989). A tutorial on Hidden Markov Models and selected applications in speech recognition. Proceedings of the IEEE, 77(2), 257-286.

Objective Questions

• Is continuous speech recognition always harder than isolated word recognition in real applications? Why?

Objective Activities

• Explore current speech recognition software tools (commercial or free ones) in languages that are familiar to you. Find out what technologies they are based on, and to what degree you think they are acceptable. Explain your reasoning, and suggest ways they might be improved.