"Detecting Repeated Patterns in Audio"

Supervised by Professor RK Moore

Abstract

The process of detecting repeating segments in audio is a fundamental part of all audio analysis in both nature and computers. Although, almost all repetitions found in natural audio are not identical, with varying substructures and spectral characteristics, and so a method with adequate tolerance to these imperfections is required. Algorithms have been presented in all fields of computer hearing with a specialization for handling particular types of audio such as music and human speech. The aim of this project is to create a flexible analysis tool capable of extracting the structure within audio samples of varying levels of complexity. The application will have the front-end functionality to modify the behaviour of existing algorithms to help us explore the performance in correctly extracting repeating segments.