Neural-network-based motion detection of a music performer from 3D position history and experimental results on an automated pure intonation musical ensemble
Kazuteru TOBITA, Ryusuke Ishikawa · AIP conference proceedings · 2024
A musical ensemble is considered one of the most advanced types of cooperative work between humans, as it is important not only to follow the musical score but also to be synchronized with other musicians, particularly their breathing.We are trying to develop an automated musical ensemble system between humans and machines, which we refer to as a "pure intonation ensemble system", as the harmonies are played in pure intonation.We employed neural-network-based motion detection and prediction for the motions of the music performer such as the "einsatz" (start of performance).The automatic performance mechanism, which accompanies the musician, then plays a frequency that is in pure-intonation harmony with the musician.