Musical Timbre Analysis

Abstract

Automatic music analysis has long been attempted, yet it is still far from achieving practically useful results. Musical audio signal consists of perceptual features such as pitch, loudness, duration and timbre, and structural aspects features such as harmony, melody, rhythm, and tempo. Complexity analysis is performed on segments of music in order to better understand the underlying generator of signals, then Independent Component Analysis was used in an attempt to extract such underlying sources. This is proved to be unsuccessful, so we investigated an alternative method of feature extraction. A simple harmonic model was used to model musical signals and an optimisation method for the parameters of the model was discussed.