"Musical Feature Extraction"

Supervised by G.Brown

Abstract

When human listeners are confronted with musical sounds, they rapidly and automatically orient themselves in the music. We are able to make numerous judgements about the composition within a few short seconds. This dissertation focuses on machine listening of music and making some of the same judgements as humans are able to.

The project consists of two separate parts. The first part examines a number of methods for the recognition of pitch and key from pieces of monophonic music producing a system capable of detecting simple musical keys from a number of songs. The second part examines methods for the extraction of beat and tempo from pieces of monophonic and polyphonic music implementing a system capable of tracking the beat and calculating the tempo from a number of different polyphonic compositions.