Extension of an audio-recordings database with features for similarity search

Art der Arbeit:

Master Thesis

Betreuer:

Meyer-Wegener, Klaus
Lehrstuhl für Informatik 6 (Datenmanagement)
Telefon +49.9131.85.27892, E-Mail: Klaus.Meyer-Wegener@fau.de

Maier, Andreas
Lehrstuhl für Informatik 5 (Mustererkennung)
Telefon +49 9131 85 27883, Fax +49 9131 85 27270, E-Mail: andreas.maier@fau.de

Beschreibung der Arbeit:

Background
The speech-analysis group of the Chair for Computer Science 5 (Pattern Recognition) has set up a database with a Web interface to record speech disorders ("PEAKS" – see http://peaks.informatik.uni-erlangen.de/ ). It has been used to analyze over 4,000 patient and control subjects so far. The database has multi-client capability, so that different institutions have their own databases, all with the same structure, but with different strategies regarding data entry. It is considered very useful to be able to search for similar recordings given a particular case.

Task
In order to be able to search for similar speech recordings, the stored recordings must be analyzed to extract appropriate features. The database schema must be extended so that these features can be stored and used for similarity search. The features may also be used in a cluster analysis. Queries for similarity search are to be defined. If suitable, index structures supporting the search should be created. The Web interface should be gracefully extended so that users can perform similarity search. A concept for multi-client search should be developed.

Vorausgesetzte Vorlesungen bzw. Kenntnisse:

Knowledge in database design; audio feature extraction

Schlagwörter:

Similarity search; audio database

Bearbeitungszustand:

Die Arbeit ist noch offen.