Similarity-based search and retrieval of audio files

Thesis Type Master
Thesis Status
Student Philip Kahle (Start: 15.01.2009) Supervisor:
Thesis Supervisor

Search parameters for audio data are currently often restricted to relate to available, textual metadata. However, there are various research projects pushing more applicable methods which allow storage, querying and retrieval especially of audio data. Alternative methods for entering queries are provided after developing new techniques for similarity analysis. This is for instance already demonstrated by existing query-by-humming systems.
This thesis shall provide a review on most recent approaches relating to audio files with a focus on those, which rely on the analysis of midi data. Afterwards, methods for optimizing the retrieval procedure are highlighted.