Prosody is an important property of spoken language. This suprasegmental information is encoded in pitch, intensity and duration patterns of syllables, words and utterances. Specific prosodic patterns (movements and values) are assumed to be used to evoke the impression of emphasised or stressed words. This property of words is called prosodic prominence. The knowledge about prosodic prominence can be used to improve both speech recognition and speech synthesis.

The aim of this bachelor thesis was to develop a tool in Python to extract 96 acoustic features related to the fundamental frequency, sound intensity and time properties of annotated speech recordings and applied it to the GRASS Corpus, a database of annotated speech recordings of 38 speakers of Austrian German. The results of this thesis indicate that duration is the most important feature to predict prosodic prominence. This tool has been used and extended in several studies within the project Cross-layer prosody modeling for conversational speech and our experiments show that the extracted acoustic features are useful to train classifiers to automatically detect prosodic prominence in read and conversational Austrian German.