Sprache der Veranstaltung:
  • Englisch
Termin Vorlesung:
  • Mi 8:15-9:45 05.025
  • 2,5 ECTS
Stunden Vorlesung:
  • 3
Stunden Übung:
  • 1
UnivIS Links:


The course concentrates on algorithms for speech and audio signal processing with applications in telecommunications and multimedia, especially

  • Human Speech Production and Hearing: Source-Filter-Model, Filterbank Model of Cochlea; Masking effects;

  • Signal Representation for Speech and Audio: Estimation and representation of short-term and long-term statistics in time domain, frequency domain, and cepstral domain; typical examples and visualisations;

  • Source Coding for Speech and Audio: criteria; scalar and vectorial quantization and coding; linear prediction; pitch prediction; waveform/ parameter/ hybrid coding; standards (ITU, GSM, ISO-MPEG)

  • Fundamental Concepts of Automatic Speech Recognition: Feature extraction, Dynamic Time Warping, Hidden Markov Models

  • Fundamental Concepts of Speech Synthesis: Text-to-Speech systems, modell-based and data-driven synthesis, PSOLA synthesis

  • Signal enhancement for acquisition and reproduction: Noise reduction, echo cancellation, dereverberation using single-channel and multichannel algorithms;

Recommendations for each topic are given during the lectures

Voraussetzungen / Organisatorisches

Signale und Systeme, Digitale Signalverarbeitung