Lecture Spracherkennung / Speech Recognition

(Here you will find binding study regulations and module handbooks. )

© TU Dresden / STKS

The lecture Speech Recognition deals with methods for converting speech audio signals into written text, i.e. "speech-to-text". After we have familiarized ourselves with the problem, there is a brief introduction to phonetics and phonology, which forms the basis for both speech recognition and speech synthesis. We then cover feature computation and single word recognition methods. We then consider the more difficult problem of word sequence recognition, using classical hidden Markov models on the one hand, and modern neural models on the other. Language models, which are used today in an extended form in ChatGPT or Gemini, also play a role in word sequence recognition.

If required, the lecture can be held in English or supported by English-language documents.

The lecture is suitable for

Also suitable for students of other technical disciplines.

Recommended knowledge: Systems theory I and II, signal processing, signal analysis and pattern recognition

The lecture is always offered in the winter semester.

Please register for participation and for further information in the corresponding Opal course.

Lecture Speech Recognition in the learning platform OPAL