Multi-scale Scattering Transform in Music Similarity Measuring

Simple item page
dc.contributor.advisorAtlas, Les
dc.contributor.authorWang, Ruobai
dc.date.accessioned2017-08-11T22:54:13Z
dc.date.available2017-08-11T22:54:13Z
dc.date.submitted2017-06
dc.descriptionThesis (Master's)--University of Washington, 2017-06
dc.description.abstractScattering transform is a Mel-frequency spectrum based, time-deformation stable method, which can be used in evaluating music similarity. Compared with Dynamic time warping, it has better performance in detecting similar audio signals under local time-frequency deformation. Multi-scale scattering means to combine scattering transforms of different window lengths. This paper argues that, multi-scale scattering transform is a good alternative of dynamic time warping in music similarity measuring. We tested the performance of multi-scale scattering transform against other popular methods, with data designed to represent different conditions.
dc.embargo.termsOpen Access
dc.format.mimetypeapplication/pdf
dc.identifier.otherWang_washington_0250O_17102.pdf
dc.identifier.urihttp://hdl.handle.net/1773/40050
dc.language.isoen_US
dc.rightsCC BY
dc.subjectmusic similarity
dc.subjectscattering transform
dc.subjecttime deformation
dc.subjectElectrical engineering
dc.subject.otherElectrical engineering
dc.titleMulti-scale Scattering Transform in Music Similarity Measuring
dc.typeThesis

Files

Original bundle

Now showing 1 - 1 of 1
Thumbnail Image
Name:
Wang_washington_0250O_17102.pdf
Size:
658.23 KB
Format:
Adobe Portable Document Format
Download

Collections

Electrical engineering