Dykiert, M; Gold, NE; (2011) Support for Learning Synthesiser Programming. In: Proceedings of the 8th Sound and Music Computing Conference (SMC 2011). Padova University Press: Padova, Italy.

Preview

Text Dykiert_Support_Learning_Synthesiser_Programming.pdf Download (190kB) | Preview

Abstract

When learning an instrument, students often like to emulate the sound and style of their favourite performers. The learning process takes many years of study and practice. In the case of synthesisers the vast parameter space involved can be daunting and unintuitive to the novice making it hard to define their desired sound and difficult to understand how it was achieved. Previous research has produced methods for automatically determining an appropriate parameter set to produce a desired sound but this can still require many parameters and does not explain or demonstrate the effect of particular parameters on the resulting sound. As a first step to solving this problem, this paper presents a new approach to searching the synthesiser parameter space to find a sound, reformulating it as a multi-objective optimisation problem (MOOP) where two competing objectives (closeness of perceived sonic match and number of parameters) are considered. As a proof-of-concept a pareto-optimal search algorithm (NSGA-II) is applied to CSound patches of varying complexity to generate a pareto-front of non-dominating (i.e. ”equally good”) solutions. The results offer insight into the extent to which the size and nature of parameter sets can be reduced whilst still retaining an acceptable degree of perceived sonic match between target and candidate sound.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item