Fakultät für Elektrotechnik und Informationstechnik
Advisor:
Schuller, Björn W. (Prof. Dr. habil.)
Referee:
Schuller, Björn W. (Prof. Dr. habil.); Bungartz, Hans-Joachim (Prof. Dr. habil.); Steinbach, Eckehard (Prof. Dr.)
Language:
en
Subject group:
DAT Datenverarbeitung, Informatik
TUM classification:
DAT 815d
Abstract:
This thesis investigates the potential of recent machine learning methods for the challenging task of information extraction from single-channel audio where the source of interest is mixed with multiple interfering sources. World-leading results are demonstrated on challenging speech separation and recognition problems where speech is mixed with non-stationary background noise such as music. Furthermore, state-of-the-art results are presented in selected music information retrieval applications involving polyphonic audio.
«
This thesis investigates the potential of recent machine learning methods for the challenging task of information extraction from single-channel audio where the source of interest is mixed with multiple interfering sources. World-leading results are demonstrated on challenging speech separation and recognition problems where speech is mixed with non-stationary background noise such as music. Furthermore, state-of-the-art results are presented in selected music information retrieval applications...
»
Translated abstract:
Diese Arbeit untersucht das Potenzial von aktuellen maschinellen Lernmethoden für das anspruchsvolle Problem der Informationsgewinnung aus einkanaligen Audiosignalen, wobei das Nutzsignal durch mehrere Störquellen überlagert ist. Weltweit führende Ergebnisse werden auf dem Problem der Trennung von Sprache und nichtstationärem Hintergrundgeräusch erzielt. Daneben werden auch mehrere Anwendungen aus der polyphonen Musikverarbeitung beispielhaft vorgestellt.