Research‎ > ‎

Electric Guitars Dataset

Content

The Electric Guitars Dataset is a database of audio recordings of electric guitars designed specifically for research on electric guitar’s sound characteristics and for the identification and/or classification of different electric guitar types. At the present moment, the dataset contains recordings of around 30 different electric guitars, and it's still being updated with plans to achieve over 50 guitars. All audio files are accompanied by parameter annotations in xml format. 


The recording setup consists of an usb audio interface connected directly to the guitar output providing the cleanest signal chain possible. The recordings are provided in mono, RIFF wave format, with 16-bit depth and 44100 Hz sampling rate. All recording sessions are performed by the same musician trying to minimize performance variations.


The database is divided into three subsets, all of them using the standard guitar tuning:

  • Subset #1 - Open Strings:

This subset contains only notes of open-strings, recorded in all pickup positions and with two different note durations, classified as long and short notes. This resulted in approximately 1400 note events.

  • Subset #2 - Chords:

This subset contains recordings of the E-Major and A-Major chords, again recorded in all pickup positions and with two different durations, totalizing 480 chords.

  • Subset #3 - A Minor Scale

The last subset is made of the notes from the A Minor Scale, starting at the 5-th fret on the low E-String, played using 3 notes per string. This subset was also recorded in all pickup positions consisting in 1620 note events.

In total, the database at its current state contains around 2000 monophonic note events and 480 polyphonic chords. The dataset is designed following some of the principles used by the Fraunhofer IDMT-SMT-Guitars dataset in order to achieve a certain degree of compatibility and will be available soon under Creative Common License.