arXiv.itsnick.co

an independent field guide to 23 million papers from arXiv & OpenAlex — not affiliated with arXiv.org

Computer Science · Signal Processing

Evaluating musical fingerprinting systems

Alastair Porter

· eScholarship@McGill (McGill)

See on the map

Abstract

Audio fingerprinting is a process that uses computers to analyse small clips of music recordings to answer a common question that people who listen to music often ask : "What is the name of that song I hear ?" Audio fingerprinting systems identify musical content in audio and search a reference database for recordings that contain the same musical features. These systems can find matching recordings even when the query has been recorded in a public space and contains added noise. Different audio fingerprinting algorithms are better at identifying different types of queries, for example, queries that are short, or have a large amount of noise present in the signal. There are few comprehensive comparisons of fingerprinting systems available in the literature that compare the retrieval accuracy offingerprinting systems with a wide range of querys.This thesis presents an overview of the historical developments in audio fingerprinting, including an analysis of three state-of-the-art audio fingerprinting algorithms. The thesis introduces factors that must be considered when performing a comparative evaluation of many fingerprinting algorithms, and presents a new evaluation framework that has been developed to address these factors. The thesis contributes the results of a large-scale comparison between three audio fingerprinting algorithms, with an analysis recommending which algorithms should be used to identify music queries recorded in different situations.

More like this