Biomechanical signal processing
Other technologies
Evidence
-
Why this level
SummaryBiomechanical signal processing has been practiced since 2016 — first at Decathlon Sports Lab on foot strike data (Kalenji, Domyos, ACAP) using Footscan/RSscan, then at ISM Protisvalor on IMU/force signals using Qualisys and Visual3D. Processing chains include stride detection, low-pass and band-pass filtering, fixed-step interpolation, phase extraction (heel strike, mid-stance, toe-off), and variable structuring for clustering or inter-group comparison. The industrialization of this chain in Python (Chaire4) and teaching it to L2 students confirm an operational and transmissible level of competency.
-
Current ceiling
SummaryReaching L4 would require publishing a versioned reusable pipeline or library, contributing in a multi-team context, and documenting methodological comparisons between tools (e.g. Qualisys vs. low-cost IMU).
-
Matlab pipeline Chaire3 — ISM 2018–2019
No sourceStride detection, filtering, interpolation to 200 points, structuring of qualitative/quantitative/time-series tables for unsupervised clustering on 47–50 runners.
-
Python interface Chaire4 — ISM 2019
No sourceIndustrialization of the chain in Python: Qualisys/Visual3D loading, heel strike, mid-stance, heel rise, toe-off detection, configurable filtering, export to .mat/.pkl/.csv.
-
Footscan data analysis — Decathlon 2016–2017
No sourceProcessing of 28,200 podological scans for the Kalenji project, foot strike and stride frequency analysis on cohorts of 248 to 1,991 runners.