Source-filter separation for bowed string instruments and its application for advanced audio effects


This project explores the possible use of source-filter separation for bowed string instruments in audio post-processing effects. Namely, two scenarios were implemented: a violin to cello and a violin to string section conversion. This is achieved by a combination of body response estimation and pitch-shifting algorithm. Special attention were given to the particular characteristics in construction and sound of bowed string instruments. The source-filter model were performed by a linear body estimation exploiting the harmonic sound nature of the strings. The pitch-shifting were implemented by a phase-vocoder approach. Furthermore, a live capable real-time implementation for Max/MSP were implemented and thoroughly tested. The resulting algorithm is practical on a consumer laptop and gives plausible results.

M.Sc. thesis, Queen Mary University of London, UK