Description

• Dataset of conductor movement during performance
• Dataset of spontaneous conducting movements
• Dataset of conducting movements performed with different articulation

• Sarasúa, Á. (2017). Musical Interaction Based on the Conductor Metaphor. Ph.D. Thesis, Universitat Pompeu Fabra.

Datasets

Conductor movement during performance

[Chapter 3]

https://repovizz.upf.edu/phenicx/datasets/

This dataset was recorded in a performance of Beethoven’s 9th Symphony played by the Orquestra Simfònica del Vallès and conducted by Rubén Gimeno. The concert took place on May 25th 2014 at the Kursaal theatre in Manresa, Spain. The recording includes data for other tasks such as automatic score alignment and source separation.

Included data:

• 32 microphones audio
• Video and audio from a videocamera capturing the whole orchestra
• From Kinect V1 facing the conductor (all at 30 fps):
  • MoCap with 3D position of nine joints: head, neck, torso, shoulders, elbows and hands.
  • RGB video (640x480 pixels).
  • RGBD video (320x240 pixels).
• Aligned score – Audio descriptors computed by Essentia

Spontaneous conducting movements

[Chapter 4]

This data was recorded in an observation study where 25 participants with different musical expertise were asked to perform spontaneous (i.e. without instructions) conducting movements on top of 3 musical fragments of a performance of Bethoven's 3rd Symphony played by the Royal Concertgebouw Orchestra.

For each participant, the dataset includes:

• Audio and scores of musical excerpts
• MoCap data captured by Kinect V1 with 3D position of fifteen joints: head, neck, torso, shoulders, elbows, hands, hips, knees and feet.
• Beat annotation (ground truth)
• Beat prediction from hand acceleration data
• Loudness prediction from body movement
• Computed MoCap descriptors

Data

Conducting movements performed with different articulation

[Chapter 5]

This data was recorded in an observation study where 24 participants with different musical expertise were asked to perform conducting movements with different articulations (legato, staccato) on top of two melodies synthesized with the corresponding articulations in 3/4 and 4/4, respectively. In Part 1 of the study, participants were not given instructions on how to perform the gestures; in Part 2, they were specifically asked to convey the different articulations through gesture variations.

For each participant, the following data is available:

• Audio of synthesized musical excerpts
• MoCap data captured by Kinect V1 with 3D position of fifteen joints: head, neck, torso, shoulders, elbows, hands, hips, knees and feet.
• Beat annotation (ground truth)
• Beat prediction from hand acceleration data
• Computed MoCap descriptors

 Data

License

These datasets are available under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.