Please use this identifier to cite or link to this item:
Title: Efficient and robust audio fingerprinting
Keywords: audio fingerprinting, spectral feature, feature extraction, fingerprint modeling, Gaussian Mixture model, string matching
Issue Date: 14-Nov-2007
Citation: FENG SHUYU (2007-11-14). Efficient and robust audio fingerprinting. ScholarBank@NUS Repository.
Abstract: As the amount of music data in multimedia databases increases rapidly, there are strong needs to investigate and develop content-based music information retrieval systems in order to support effective and efficient analysis, retrieval and management for music data. This thesis examines the problem of content-based music identification by efficient and robust audio fingerprinting. Firstly, we study and compare several spectral features, and find that product spectrum based feature which combines phase spectrum with magnitude spectrum is more robust than other spectral features which are derived only from magnitude spectrum. Secondly, a pattern accumulative similarity is proposed to better measure the similarity between audios under several types of distortions. Thirdly, Gaussian mixture model is used to model audio fingerprints, boosting the robustness of audio fingerprints under noise distortions while making fingerprints more concise. Compared with existing approaches, our method is more resistant to distortions.
Appears in Collections:Master's Theses (Open)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
FengSY.pdf591.81 kBAdobe PDF



Page view(s)

checked on Oct 7, 2020


checked on Oct 7, 2020

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.