n the past few years, thanks to the increasing avail- ability of multimedia sharing platforms, the online availability of user generated content has incredibly grown. However, since media sharing is often not well regulated, copyright infringement cases may occur. One classic example is the pirate distribution of audio bootlegs, i.e., concerts illegally recorded using portable devices. In order to guarantee copyrights and avoid the sharing of such illicit material, in this paper we propose an automatic audio bootleg detector. This can be used to analyze audio data in bulk, in order to filter out from a database the audio tracks recorded, e.g., by fans during a live performance. To this purpose, we propose to use a set of acoustic features to characterize audio bootlegs, justified by theoretical foundations. Then, we train a bi- nary classifier that operates on this set of features to discriminate between: i) audio tracks recorded at either concerts, clubs, or theaters; ii) legally distributed live performances professionally mixed and edited. In order to validate our system, we tested it on a dataset of more than 250 audio excerpts considering different musical genres and different kinds of music performances. The results achieved are promising, showing a high bootleg detection accuracy.

Feature-based classification for audio bootlegs detection

BESTAGINI, PAOLO;ZANONI, MASSIMILIANO;SARTI, AUGUSTO;TUBARO, STEFANO
2013

Abstract

n the past few years, thanks to the increasing avail- ability of multimedia sharing platforms, the online availability of user generated content has incredibly grown. However, since media sharing is often not well regulated, copyright infringement cases may occur. One classic example is the pirate distribution of audio bootlegs, i.e., concerts illegally recorded using portable devices. In order to guarantee copyrights and avoid the sharing of such illicit material, in this paper we propose an automatic audio bootleg detector. This can be used to analyze audio data in bulk, in order to filter out from a database the audio tracks recorded, e.g., by fans during a live performance. To this purpose, we propose to use a set of acoustic features to characterize audio bootlegs, justified by theoretical foundations. Then, we train a bi- nary classifier that operates on this set of features to discriminate between: i) audio tracks recorded at either concerts, clubs, or theaters; ii) legally distributed live performances professionally mixed and edited. In order to validate our system, we tested it on a dataset of more than 250 audio excerpts considering different musical genres and different kinds of music performances. The results achieved are promising, showing a high bootleg detection accuracy.
2013 IEEE International Workshop on Information Forensics and Security (WIFS)
9781467355933
File in questo prodotto:
File Dimensione Formato  
wifs13_submission_70_camera_ready.pdf

Accesso riservato

: Altro materiale allegato
Dimensione 298.8 kB
Formato Adobe PDF
298.8 kB Adobe PDF   Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11311/821730
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 6
  • ???jsp.display-item.citation.isi??? 0
social impact