Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Detecting phonemes within the singing of polyphonic music

: Gruhne, Matthias; Schmidt, Konstantin; Dittmar, Christian

Fulltext (PDF; )

Schubert, E. ; ARC Research Network in Human Communication Science -HCSNet-, Sydney:
Inaugural International Conference on Music Communication Science, ICoMCS 2007. Proceedings. CD-ROM : University of New South Wales, Sydney, Australia, December 5 to 7, 2007
Sydney: HCSNet, 2007
ISBN: 978-1-74108-161-9
International Conference on Music Communication Science (ICoMCS) <2007, Sydney>
Conference Paper, Electronic Publication
Fraunhofer IDMT ()
vocal analysis; phoneme detection

Automated detection of phonemes in polyphonic music is an important prerequisite for synchronizing music and corresponding lyrics. This paper describes a novel approach of automatic phoneme estimation within digitized music pieces. Since there are already a number of publications, aiming at distinguishing singing passages and non-singing passages in music, this paper only concentrates on detecting voiced phonemes in singing passages of music. In a first step, the leading melody of a segment is recognized, subsequently the harmonics are extracted and selected. Thereafter, the harmonics are resynthesized to an audible signal. Finally, a common feature extraction and classification algorithm is applied using a number of different classifiers.