Novelty measures as cues for temporal salience in audio similarity

Mark Cartwright*, Bryan A Pardo

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Scopus citations

Abstract

Most algorithms for estimating audio similarity either completely disregard time or they treat each moment in time equally. However, many studies over the years have noted several factors that affect how much attention we give to certain sounds or parts of sounds (e.g. loudness, the attack, novelty). These findings suggest that some time segments of audio may be more salient than others when making similarity judgments. We believe that if we could estimate this information, we could improve audio similarity measures. This paper presents the results of a human subject study designed to test the hypothesis that sounds segments with high timbral change are more salient than segments with low timbral change. We then investigate whether we can use this information to improve two audio similarity measures: a "bag-of-frames" approach and a dynamic time warping approach.

Original languageEnglish (US)
Title of host publicationMIRUM 2012 - Proceedings of the 2nd International ACM Workshop on Music Information Retrieval with User-Centered and Multimodal Strategies, Co-located with ACM Multimedia 2012
Pages51-56
Number of pages6
DOIs
StatePublished - Dec 10 2012
Event2nd International ACM Workshop on Music Information Retrieval with User-Centered and Multimodal Strategies, MIRUM 2012 - Co-located with ACM Multimedia 2012 - Nara, Japan
Duration: Nov 2 2012Nov 2 2012

Publication series

NameMIRUM 2012 - Proceedings of the 2nd International ACM Workshop on Music Information Retrieval with User-Centered and Multimodal Strategies, Co-located with ACM Multimedia 2012

Other

Other2nd International ACM Workshop on Music Information Retrieval with User-Centered and Multimodal Strategies, MIRUM 2012 - Co-located with ACM Multimedia 2012
CountryJapan
CityNara
Period11/2/1211/2/12

Keywords

  • Audio novelty measures
  • Audio similarity
  • Audio temporal salience
  • Query-by-example

ASJC Scopus subject areas

  • Music
  • Information Systems

Fingerprint Dive into the research topics of 'Novelty measures as cues for temporal salience in audio similarity'. Together they form a unique fingerprint.

Cite this