"Acoustic Music-based Pose Learning (AMPL)"

"The Acoustic Music-based Pose Learning (AMPL) dataset is a novel collection linking musical sensing signals with 3D human poses. It comprises over 4 hours of synchronized acoustic and motion data, consisting of 27 recording sessions from 9 subjects. The dataset utilizes four different background music (BGM) tracks spanning two genres\u2014Ambient and Jazz\u2014as sensing signals. All data were captured in an ordinary classroom environment featuring realistic background noise and reverberation. The technical configuration consists of high-resolution commercial speakers and an ambisonics microphone placed one meter from subjects in a line-of-sight setup. Precise ground-truth 3D poses were obtained via a 16-camera motion capture system. Additionally, to enable the evaluation of sound absorption and system robustness, the dataset includes subjects wearing both specialized motion capture suits and everyday plain clothes."