Recognising Cello Performers Using Timbre Models

In this paper, we compare timbre features of various cello performers playing the same instrument in solo cello recordings. Using an automatic feature extraction framework, we investigate the differences in sound quality of the players. The motivation for this study comes from the fact that the performer’s influence on acoustical characteristics is rarely considered when analysing audio recordings of various instruments. While even a trained musician cannot entirely change the way an instrument sounds, he is still able to modulate its sound properties obtaining a variety of individual sound colours according to his playing skills and musical expressiveness. This phenomenon, known amongst musicians as “player timbre”, enables to differentiate one player from another when they perform an identical piece of music on the same instrument. To address this problem, we analyse sets of spectral features extracted from cello recordings of five players and model timbre characteristics of each performer. The proposed features include harmonic and noise (residual) spectra, Mel-frequency spectra and Mel-frequency cepstral coefficients (MFCCs). Classifiers such as k-Nearest Neighbours (k-NN) and Linear Discrimination Analysis (LDA) trained on these models are able to distinguish the five performers with high accuracy. Magdalena Chudy Centre for Digital Music, Queen Mary University of London Mile End Road, London E1 4NS, United Kingdom e-mail: [email protected] Simon Dixon Centre for Digital Music, Queen Mary University of London Mile End Road, London E1 4NS, United Kingdom e-mail: [email protected]