General Information

This lecture course addresses the issue of automatic speech recognition, i.e. mapping an acoustic signal to a sequence of words. First various concepts of pattern recognition are introduced including hidden Markov models and language models built from large corpora of (symbolic) text. The course Speech Communication I is not required for this course.

The current lecture material can be found in the TeachCenter.

Contents

• Introduction to Automatic speech recognition (ASR)
• Speech Production & Phonetics
• Feature extraction
• Classification
• Estimation of probability distributions
• Gaussian Mixture Models
• Markov models
• Hidden Markov models (HMMs)
• Grammar models - Language Models
• Decoding (Viterbi decoder)
• Deep Neural Networks for Speech Recognition

Lecture notes (old slides)

• Course overview/Introduction to ASR
• Classification
• Hidden Markov Models
• Duration modeling in HMMs
• Acoustic/phonetic Elements
• Grammar Models
• Word Sequence Decoding
• Sinusoidal Modeling/Harmonic-plus-Noise Model

References

Speech recognition

• D. Jurafsky et al: Speech and Language Processing: An Introduction to Natural Language Processing, Computational Linguistics and Speech Recognition. Prentice-Hall 2009.
• X. Huang, A. Acero, H.-W. Hon: Spoken Language Processing: A Guide to Theory, Algorithm and System Development. Prentice Hall PTR 2001.

Classification

• R.O. Duda and P.E. Hart: Pattern Classification and Scene Analysis. Wiley and Sons, Inc., 1973.
• C. M. Bishop: Pattern Recognition and Machine Learning, Springer, 2006