Automatic Annotation of Timbre Variation for Musical Instruments

This paper proposes a preprocessing technique for the automatic transcription of performances produced by a musical instrument (or other sound source) capable of timbre variations. Voice recognition techniques will be exploited to gather information about timbre, then a clustering approach will be used to reduce data cardinality, and, finally, data dimensionality will be further reduced using multi-dimensional scaling to create labels as points in a data-driven timbre-space. A graphical visualization of the achieved results will be implemented in order to verify the achievement of the initial requirements. A MATLAB toolkit performing the operations described in this paper is publicly available to test the effectiveness of the proposed approach.