Musicians are able to recognise the subtle differences in timbre produced by different playing techniques on an instrument, yet there has been little research into achieving this with a computer. This paper will demonstrate an automatic system that can successfully recognise different timbres produced by different performance techniques and classify them using signal processing and classification tools. Success rates over 90% are achieved when classifying snare drum timbres produced by different playing techniques.
Adam R. Tindale, Ajay Kapur, George Tzanetakis, Ic