Parkinson's disease alters gait and bilateral coordination, and the performance of machine‑learning classifiers heavily depends on how continuous gait signals are represented. This work investigates whether preserving anterior‑posterior center‑of‑pressure (AP‑COP) information at fixed locations across a normalized stance provides a compact and informative representation of plantar vertical ground reaction force (VGRF) signals.
Bilateral vertical ground reaction force recordings from 165 participants in the Gait in Parkinson's Disease Database were evaluated using repeated fully nested participant‑level cross‑validation. We propose AP‑COP10, which consists of AP‑COP position and bilateral asymmetry across five stance windows, yielding ten features. AP‑COP10 achieved an AUC of 0.894, outperforming three harmonized literature‑derived COP representations under the same evaluation pipeline.
A complementary set of 25 non‑AP‑COP descriptors alone reached an AUC of 0.856, while the full 35‑feature representation achieved 0.908. Removing AP‑COP10 from the full set caused a statistically significant drop in discrimination, whereas adding the complementary descriptors to AP‑COP10 produced only a small, non‑significant improvement. Feature competition indicated that the most informative stance‑indexed descriptors were concentrated in early and early‑mid stance phases. Source‑study holdout and sensor‑perturbation analyses further supported the robustness of the representation.
These findings demonstrate that stance‑indexed AP‑COP retains discriminative information not readily recovered by broader engineered gait descriptors, supporting compact and interpretable representations for machine‑learning analysis of pathological gait.
Review