Musical understanding for music information retrieval
In our time music culture is largely computer-based regarding listening, production and distribution. This triggers a need for musically intelligent tools that understand music in a manner similar to the way in which humans recognise e.g., melody, meter or expressive character not least in the new expanding field of Music Information Retrieval, MIR. Since research on MIR has been directed mainly by experts in computer modelling, the musicological basis is often weak limiting the results. This project combines competence in music technology, cognition and musicology and takes its starting point in Ahlbäck’s new music analysis model and the KTH model of music performance of Friberg et al. We will develop new computational models for music analysis based on cognitive and perceptual theories, extending Ahlbäck’s model, the stylistically most general model of melody structure at present. It will include analysis of polyphony, phenomenal accent and tonality, as well as predicting high-level semantic musical descriptors. A demonstrator will be constructed which makes predictions of emotional expression of music, allowing “Musical mood search”. Further applications include intelligent music search, automatic transcription and tools for comparative music studies. It will require basic research in music – such as the relationship between surface structure and tonality cognition in a cross-cultural perspective – and thus provide new insights in music communication and understanding.
The work is performed in co-operation with Dr. Sven Ahlbäck at the Royal College of Music in Stockholm, Department of Folk Music.
Group: Sound and Music Computing
Anders Friberg (Project leader)
Duration: 2006-01-01 - 2009-12-31
KTH research database: http://researchprojects.kth.se/index.php/kb_1/io_9745/io.html
Keywords: Computer modelling, Emotions, Music analysis, Music cognition, Music information retrieval
Journal of New Music Research, 38(2), 155-169. [pdf] (2009). Recognition of the main melody in a polyphonic symbolic score using perceptual knowledge.
Proc. of the 11th Int. Conference on Digital Audio Effects (DAFx-08), Espoo, Finland (pp. 1-6). [pdf]
(2008). . In