Course content

Acoustic description of speech production. Digital models for production of speech. Physiology of the ear and hearing. What we hear and what we perceive: human perception of speech and sound. Linguistic and statistical description of the speech signal. Methods for speech analysis. Basic methods for statistical pattern recognition. Automatic speech recognition with emphasis on statistical methods (Hidden Markov Models): statistical methods for acoustic and linguistic modeling; principles for efficient decoding (recognition). Speech synthesis, speech concatenation and text-to-speech synthesis: Text analysis, prosodic modelling and sound synthesis.

Learning outcome

The course aims at giving an understanding of basic properties of speech, speech production and perception as well as an insight in how this understanding can be applied for constructing systems for automatic speech recognition and speech synthesis.

Learning methods and activities

Lectures, voluntary homework problems and mandatory group based computer excercises. If there is a re-sit examination, the examination form may be changed from written to oral.

Compulsory assignments

• Exercises

Recommended previous knowledge

TTT4120 Digital Signal Processing or equivalent.

Course materials

Text book will be announced at semester start.

Credit reductions