JNMR Tabla DB

The JNMR tabla database is comprised of 35 compositions containing a total of 27,000 strokes. The symbolic notation (in a Humdrum format called **bol) and corresponding audio are given. For further information regarding the encoding scheme please see: Parag Chordia. "Automatic transcription and representation of solo tabla music." Computing in Musicology. Vol. 13. The audio was synthesized using professionally recorded samples. For each stroke category, there are approximately ten samples, which are chosen randomly to simulate the natural variance in timbre between strokes of the same category. Please reference the following publication when using this database:
P. Chordia, A. Sastry, and S. Senturk. Predictive tabla modeling using variable-length markov and hidden markov models. Journal of New Music Research, 40(2):105-118, 2011.

idm09 (tempo db)

The idm09 used in our ISMIR 2009 paper consists of tenty-seven 30 second excerpts chosen from the IDM/glitch genre of electronic music, with an emphasis on tracks that we thought were rhythmically complex. For each excerpt, two independent manual annotations were made.

A compressed, zipped tarball of all the mp3 files: idm09.tar.gz

Tempo ground truth (textfile): idm09ground.txt

The excerpts can be previewed individually here

nicm08

We hope to make this database (feature matrices and recordings) available for download soon.

This database was initially assembled for a project in applying content-based recommendation techniques to Indian Classical music. For this reason its content is quite wide ranging, in terms of sub-genre, era, recording quality, performer, instrument, etc.

GTraagDB

This database was assembled primarily from recordings that I made in Kolkata in 2006. Most of the recordings contain only the solo instrument or voice. The recordings were made with accompanying instruments, mainly tanpura, and tabla for metric sections, but the solo instrument was extracted from the mutli-tracked sessions. These recordings are excellent for melody classification and modeling tasks. An accompanying text file gives the artist, raag, instrument, and tonic frequency ("Sa").