Nick Collins

Home [Music] Research {Software} [Teaching] Contact


To accompany the paper:

Nick Collins, Peter Manning and Simone Tarsitani (2018) "A new curated corpus of historical electronic music: Collation, data and research findings". Transactions of ISMIR


[Metadata]: Tab-separated text file, one piece per line

Feature extraction data is provided as binary archives (intended for SuperCollider), and as ascii text files. Two feature extraction runs are supplied, the first a custom set of 22 features as averages across pieces and window by window, as described in the paper, and the second a set of basic frame by frame MFCCs and 12 steps per octave chroma as requested by an anonymous paper reviewer.

[Feature extraction data]: 22 features, max-min normalized, SuperCollider archive format (173 MB)

[SuperCollider code] to load feature data from binary archives directly to SuperCollider, and illustrating reading metadata file

[Feature extraction data]: 22 features, max-min normalized, ascii text file format (337 MB)

[SuperCollider code] to extract feature data given the original audio (audio not supplied, but code illustrates exact open source code location/definition of each feature)

[StereoSpatialEbb SuperCollider code] Plug-in for additional feature extractor used for amount of stereo spatial movement

Feature mins and maxs across data used for normalisation (also supplied with download sets above):

[ 0.19248595833778, 0.0, 0.0, 0.0, -7.1525573730469e-07, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.44318243861198, 0.67681056261063, -3.3816486393334e-05, -0.00038114056224003, 0.0, 0.0, -0.0003555714210961, -0.00016408874944318, -0.00027990210219286, 0.0, 0.0 ]

[ 81.143112182617, 0.86279946565628, 19521.849609375, 16.831871032715, 0.81514406204224, 189.56230163574, 31.0, 1.9969160556793, 0.92879813909531, 6.6438584327698, 2.7908728122711, 1.0, 1.0, 0.82183212041855, 0.74596321582794, 6.7707462310791, 539.63818359375, 1.3394432067871, 0.67628860473633, 0.99919664859772, 77.468315124512, 55.247421264648 ]

[12 MFCCs and (12TET) Chroma]: unnormalized, ascii text file format, one text file per audio file (frame by frame, around 43Hz, 1024 samples hop size within 44100 sampling rate) (4.7GB)