An AutismAI Lab application

Objective motor phenotyping, in the browser.

Volitiq turns any device with a camera into a research-grade motor assessment tool. It captures skeletal landmarks on-device with MediaPipe,[Lugaresi et al., 2019] computes interpretable kinematic features, and lets researchers and clinicians track how people move over months and years — with no installation, no specialist hardware, and raw video staying on the device by default.

Browser-native
The webcam is the sensor

MediaPipe extracts face, body and hand landmarks at ~30 fps directly in the browser, on consumer laptops, tablets and phones — no plug-in, no lab.

Interpretable
40+ features, 7 domains

Every recording yields 40+ named kinematic features across seven motor-control domains — each with a definition and a formula, not a black-box score.

Cross-population
One pipeline, many cohorts

The same feature framework applies to autism research, physiotherapy progress, elderly mobility and fall risk, and inclusive education.

The challenge

Motor differences are well documented — but measuring them objectively is gated behind specialist labs.

Motor coordination differences appear in a large share of autistic individuals — meta-analysis puts the effect size well above the threshold for clinical significance.[Fournier et al., 2010] The "micro-movement" perspective argues these fine-grained motor signals carry information that standard observation misses.[Torres et al., 2013]

Yet objective motor assessment has traditionally required force transducers or marker-based motion capture, confined to a small set of research clinics. The same gating applies to aging, rehabilitation and disability research. Volitiq removes the apparatus: the skeletal-pose features that drive modern motor research are recoverable from ordinary webcam video.

What Volitiq measures

Seven motor-control domains.

The 40+ features are organised into seven domains grounded in sensorimotor-control research.[Wolpert et al., 2011] Movement smoothness, for example, is quantified with established, validated measures.[Balasubramanian et al., 2015]

01 Initiation

How movement starts — reaction time and time-to-peak-velocity from the moment a task begins.

02 Smoothness

How fluid a movement is, via jerk-based measures and spectral arc length (SPARC).

03 Variability

Trial-to-trial consistency of a repeated movement — the stability of a motor pattern.

04 Coordination

Phase relationships between joints and limbs, recovered with Hilbert-transform phase coupling.

05 Rhythm

The timing regularity of repetitive or cyclic movement.

06 Complexity

Signal complexity of the movement trace, via approximate entropy and fractal dimension.

07 Efficiency

Economy of movement — path efficiency between start and target.

How the analysis works

Transparent statistics, not a black box.

From the landmark stream, Volitiq computes velocity, acceleration and jerk with central-difference derivatives and Savitzky-Golay smoothing, then derives the domain features. Group comparisons use classical, inspectable statistics — every number can be traced back to a formula.

Welch's t-test

Unequal-variance group comparison across features.

Cohen's d

Effect size, so differences are reported by magnitude, not just significance.

Feature importance

A transparent ranking that combines significance and effect size.

Spectral (FFT)

Frequency-domain analysis for rhythm and longitudinal change.

Machine learning: bring your own model

Volitiq is a capture-and-analysis platform. It does not ship trained diagnostic models. Research teams that have validated their own classifiers can apply them to new recordings through the platform; the classifier, its training data, and its claims remain the researcher's own.

Tracking change over time

Three dashboards, one secure backend.

Longitudinal

Per-participant, session-over-session views of how velocity, acceleration, jerk and derived features change across weeks and months.

Inter-group

Cohort comparisons with statistical effect sizes, so a group difference is quantified, not just visualised.

Intra-group

Distribution and correlation views within a single group, for exploratory analysis before a study is scaled.

Studies are organised as user → studies → participants → sessions, with per-study data isolation enforced at the storage layer. By default only derived landmarks and features are stored; raw video stays on the device.

Under the hood

Markerless pose capture, in the browser.

Volitiq captures with MediaPipe Tasks Vision — face-mesh, body-pose and hand-tracking models running client-side on the GPU.[Lugaresi et al., 2019] Markerless pose estimation from ordinary video is the same research direction pioneered by systems like OpenPose,[Cao et al., 2019] now efficient enough to run live in a web page.

Guided recording gives non-expert operators real-time distance and lighting feedback, so a caregiver or teacher can collect usable data without training. A serverless AWS backend handles feature extraction and storage, keeping the running cost low enough for population-scale studies.

What Volitiq does not claim

A research platform, not a diagnostic medical device.

Volitiq began as the autism-focused E-MotionSpec research project and now generalises across populations. It produces research-grade measurements — nothing more, and nothing hidden.

Not a diagnostic instrument

Volitiq is a research platform, not a medical device, and does not provide a diagnostic claim about any condition.

Not a motion-capture replacement

For sub-millimetre precision, marker-based motion capture remains the gold standard. Volitiq optimises for accessibility at research-useful accuracy.

Not an autonomous classifier

The platform surfaces features and statistics. Any predictive model is one a research team supplies and validates itself.

Sources

References

External references are peer-reviewed. The lab's own manuscripts on this work are listed by status on the science page and the Publications page.

  1. Fournier, K.A., Hass, C.J., Naik, S.K., Lodha, N., & Cauraugh, J.H. (2010). Motor Coordination in Autism Spectrum Disorders: A Synthesis and Meta-Analysis. Journal of Autism and Developmental Disorders 40(10), 1227–1240. doi:10.1007/s10803-010-0981-3
  2. Torres, E.B., Brincker, M., Isenhower, R.W., Yanovich, P., Stigler, K.A., Nurnberger, J.I., et al. (2013). Autism: The Micro-Movement Perspective. Frontiers in Integrative Neuroscience 7, 32. doi:10.3389/fnint.2013.00032
  3. Wolpert, D.M., Diedrichsen, J., & Flanagan, J.R. (2011). Principles of Sensorimotor Learning. Nature Reviews Neuroscience 12(12), 739–751. doi:10.1038/nrn3112
  4. Balasubramanian, S., Melendez-Calderon, A., Roby-Brami, A., & Burdet, E. (2015). On the Analysis of Movement Smoothness. Journal of NeuroEngineering and Rehabilitation 12, 112. doi:10.1186/s12984-015-0090-9
  5. Lugaresi, C., Tang, J., Nash, H., McClanahan, C., Uboweja, E., Hays, M., et al. (2019). MediaPipe: A Framework for Building Perception Pipelines. arXiv:1906.08172
  6. Cao, Z., Hidalgo, G., Simon, T., Wei, S.-E., & Sheikh, Y. (2019). OpenPose: Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields. IEEE Transactions on Pattern Analysis and Machine Intelligence 43(1), 172–186. doi:10.1109/TPAMI.2019.2929257

Try Volitiq, or bring it to your cohort.

Researchers can launch the app and capture a session in minutes. Labs and clinics interested in deploying Volitiq across a study cohort can reach the team through the collaboration page.