QUARTET dataset

The QUARTET dataset contains 96 multimodal recordings of string quartet performance. Acquired data includes the audio of each performance in separate tracks: up to two ambient tracks and individual tracks acquired through contact microphones.

In addition, each recording includes instrumental motion capture and a set of derived instrumental bowing descriptors for each performer, upper body movement for each performer, as well as high-quality video.

The recordings were carried out as part of the experiments on ensemble performance interdependence reported in my doctoral dissertation titled A computational approach to studying interdependence in string quartet performance.