Relative Contributions of Body Geometry and Movement Style

Determine the respective contributions of body stature and personal movement style to identity information in millimeter-wave point-cloud recordings of activities of daily living, which the reported component ablations do not isolate.

Background

The study uses a dual-stream identity architecture that combines time-aggregated spatial structure with frame-to-frame dynamics. Although this design captures cues plausibly associated with body geometry and movement patterns, the component ablations compare static and dynamic representation streams without independently identifying which physical or behavioral sources produce the learned identity information.

Consequently, the paper leaves unresolved whether recognition performance is driven primarily by relatively stable characteristics such as stature and body build, by personal movement style, or by their interaction. Resolving this issue would clarify the biometric basis of identification across the seven activities of daily living and help assess how well the method may generalize beyond the controlled collection protocol.