Back to list

Extraction of Prosodic Features from Speech Signals

Authors Pokorny, F.
Year 2011
Thesis Type Audio Engineering project
Topic Audio Signal Processing
Keywords speech processing
Abstract Human speech carries two types of information. On the one hand, semantics covers the content of an utterance, on the other hand prosody provides the expression’s structure and accentuation. Thus, speech features are termed prosodic features if their direct segmentation from the syntagmatic run is not possible and a comparison with previous and/or following segments of the signal is required. So the relation between these segments is essential for the representation of a prosodic linguistic function. The prosodic features intonation and intensity are characterized by the acoustic parameters fundamental frequency and sound intensity. Speech rhythm can be extracted by detecting the borders of syllables. The aim of this project was the development of an analysis/synthesis tool in MATLAB and/or Praat that extracts prosodic features from a speech signal and, furthermore, creates a synthetic signal consisting of these features only. Moreover, they are scalable using a GUI before the synthesis.
Supervisors Jany-Luig, J.