Layer 2 – Low-Level Features
Low-level features are instantaneous descriptors of movement, usually computed directly from raw data (Layer 1) or from short sliding windows of samples.
They are typically represented as time-series with the same sampling rate as the input signals.
Examples of Low-Level Features
| Feature | Description | Implemented |
|---|---|---|
| Kinematics 1 | Velocity, acceleration, and jerk. | |
| Kinetic Energy 1 | Energy of a cloud of 3D moving joints, possibly weighted by their masses using biometric tables. | |
| Motion Index / Quantity of Motion (QoM) 1 | Area of the difference between silhouettes in consecutive frames. | |
| Postural Contraction 1 | Extent to which body posture is close to its barycenter. | |
| Smoothness 1 | Motion of a joint according to biomechanics laws of smoothness. | |
| Postural Balance 1 | Projection of the body’s barycenter onto the floor within the support area of the feet | |
| Postural Tension 1 | Vector describing angular relations between feet, hips, trunk, shoulders, and head; inspired by angles in classical painting/sculpture used to express tension. |
References
-
Camurri, Volpe, Piana, Mancini, Alborno, Ghisio (2018) The Energy Lift: automated measurement of postural tension and energy transmission. Proc. MOCO 2018 ↩↩↩↩↩↩↩
-
Camurri, Volpe, Piana, Mancini, Niewiadomski, Ferrari, Canepa (2016) The Dancer in the Eye: Towards a Multi-Layered Computational Framework of Qualities in Movement, Proc. MOCO 2016. ↩