HTTP/2 200 date: Sat, 17 Jan 2026 22:41:19 GMT server: Apache/2.4.66 (Debian) last-modified: Mon, 24 Dec 2012 23:35:31 GMT etag: "89a-4d1a1a724f6c0-gzip" vary: Accept-Encoding content-encoding: gzip content-length: 1092 content-type: text/html; charset=utf-8 x-varnish: 221320769 age: 0 via: 1.1 pannekake.samfundet.no (Varnish/7.7) accept-ranges: bytes

Musical descriptors

An assessment of psychoacoustical models in the presence of lossy compression

Not much information here, but at least you can download:

• A PDF version (1.4MB, PDF) of my master's thesis, exactly as it was sent to printing (which is the reason why it's missing a front page and problem description -- that was added automatically from NTNU; it is also the reason why the margins may look a bit odd).
• A electronic copy of the source code for the feature extraction and the Voronoi searcher, as produced in the back of the report (with a few extra build scripts etc.), in .tar.bz2 or .zip format. Together with the data below, this should be enough to verify all the data material in the report. Note that the build scripts for that are rather crude, though :-)
• A copy of the extracted feature data (153MB, tar.bz2) from all tracks used, along with a file list with duplicates removed, used by the build scripts. Most people really do not want all this data.