Gait Biometrics Lab
Recognition and movement signatures
An environment for exploring gait as a biometric: how a stride becomes movement features, how features become a signature, and what covariates and privacy do to that signature. Grounded in the published recognition work, with no result stated ahead of the evidence.
GaitAI Labs states only what the published record supports. No dataset figure, recognition result or accuracy appears here until it has been released and can be cited.
From a stride to a signature, module by module.
Six modules in pipeline order. Each states what it is for and what is true of it today: 3 of the 6 have something that runs now, elsewhere on the site, and are linked to it. The rest are the framework the lab grows into.
Stride capture
A walk, recorded from the side or from a camera in place, segmented into strides at the canonical gait events — heel strike, loading, mid-stance, toe-off, swing.
Runs todayPose landmarks and the body's path through the frame are extracted from a real clip today, in the browser, by the Movement Intelligence Lab's analyzer.
Run the analyzer on a clipMovement features
The stride as temporal channels — joint trajectories, ground contact, vertical travel, cadence-bearing rhythms — the Motion DNA a signature is built from.
Runs todayThe analyzer derives Motion DNA channels from a clip today. They are shown as channels, not scored.
See Motion DNA on a clipSignature
The channels reduced to one compact, person-specific representation — the movement signature that gait recognition compares.
Module to followModule to follow. No signature is computed on this site and no representation is shown as a result.
Covariates
What changes the walk without changing the walker: viewing angle, clothing, carried objects, speed and surface. Robustness to them is the central problem of the published work.
Module to followModule to follow, grounded in the covariate-invariant recognition papers listed below.
Privacy
How a gait dataset and a signature are protected inside a deep-learning pipeline, and what an privacy-aware path keeps versus gives up.
Runs todayThe Privacy Lens in the Movement Intelligence Lab walks the privacy-aware path today, on illustrative data. The dataset-protection work is published.
Step through the Privacy LensMatching
Comparing one signature against enrolled ones — the recognition step, and the one where every claim must carry its evidence.
Module to followModule to follow. Nothing on this site performs recognition, and no accuracy is stated anywhere in GaitAI Labs.
7 surfaced academic papers on gait recognition.
Deep-learning gait recognition with covariates, model-based recognition, pattern recognition for gait, and recognition from pose features. These are the record; the lab's modules are built against them and claim nothing beyond them.
- 2023Deep learning pipelines for recognition of gait biometrics with covariates: a comprehensive reviewArtificial Intelligence Review
- 2022Intra-class variations with deep learning-based gait analysis: A comprehensive survey of covariates and methodsNeurocomputing
- 2024Advancements in Artificial Intelligence for Biometrics: A Deep Dive into Model-based Gait Recognition TechniquesEngineering Applications of Artificial Intelligence
- 2024Journey into Gait Biometrics: Integrating Deep Learning for Enhanced Pattern RecognitionDigital Signal Processing
- 2023Data Preprocessing and Feature Selection Techniques in Gait Recognition: A Comparative Study of Machine Learning and Deep Learning ApproachesPattern Recognition Letters
- 2023Comparative Study of Machine Learning and Deep Learning Techniques for Gait Recognition: Advances in Feature Extraction, Reduction, Transformation, and ClassificationImage and Vision Computing
- 2022A robust covariate-invariant gait recognition based on pose featuresIET Biometrics