Output details
13 - Electrical and Electronic Engineering, Metallurgy and Materials
Queen Mary University of London : A - Electrical and Electronic engineering
Generating and Evaluating Musical Harmonizations That Emulate Style
Novel framework combined statistical learning and music-theoretic approach to learn harmonization style from sparse input, and provided quantitative evaluation. Raised issue: why less data may be better for style specificity. Resulting harmonisations less generic than Microsoft's MySong. Led to student winning 2010 Grace Hopper Best New Investigator Paper Award. Generated press (Wired, Telegraph, MSNBC, USA Today, Fox News) and coverage in IEEE Intelligent Systems. Culmination of highly cited 'Hybrid System for Automatic Generation of Style-Specific Accompaniment' (Computational Creativity Workshop), 'Evaluating Style-Emulation Effectiveness', 'Interactive Decision-Support Tool' (International Conference on Music Information Retrieval). CMJ (MIT Press) is leading journal in computer music.