We present the spectral analysis of conducting physical movements by exploiting
machine learning systems pre-trained human pose vector representations. 2 or 3-
dimensional dynamic motions are extracted from 2D original video recorded in
actual musical live performances. The fundamental movement of conducting are
projected into angular velocity and angular acceleration according to the Angular-
Dynamics method originated with Pierre Boulez and Ken Ito. Musculoskeletal
motion spectrum are calculated when Fourier transform is applied. No previous
work discussing spectrums and harmonics of corporal musical movements exists. A
new way of systematization on various musical performances would be expected
from here.