Evaluating Motion Capture Systems using the STRN Quality Framework

When it comes to motion capture, accuracy isn’t optional — it’s the foundation of trust in your data.

In practice, “accuracy” depends on more than the system. It’s shaped by protocols, models, and who’s collecting the data. Which is exactly why it needs a framework. 

In 2023, the Sports Technology Research Network (STRN), in collaboration with 48 leading experts across science, engineering, and performance, released a Quality Framework for Sports Technologies. It outlines 25 measurable features, grouped under five pillars, to help researchers, labs, and technology buyers systematically assess motion capture solutions. 

In this five-part series, we’re unpacking what each of those pillars actually means in practice, starting with Pillar A: Quality Assurance & Measurement, and why it’s more than just a spec sheet.

We’ll dig into:

• What “accuracy” actually means in a biomechanics context (and where vendors oversimplify it)
• Why repeatability ≠ reproducibility — and why you need both
• How to spot red flags when evaluating vendor claims
• The limitations of “gold standard” tools like optical motion capture
• Where markerless systems like Theia3D are closing the gap
  ‍

Why Accuracy is More Complicated Than You Think

Accuracy is one of the first things people ask about when evaluating a motion capture system, and rightly so. But the definition of accuracy in biomechanics isn’t always straightforward.

At its core, accuracy refers to how closely a system’s output matches a trusted reference — usually, optical marker-based motion capture. That’s the benchmark most researchers and performance professionals are familiar with. But it comes with its own flaws: operator sensitivities, soft tissue artefacts, and inconsistent model definitions all contribute to known inaccuracies in marker-based data.

So the real question becomes: accurate compared to what?

Modern systems like Theia3D are often benchmarked against these traditional setups, and in many cases, perform within the same error ranges (or better), depending on the application. Here’s what to look for:

• Joint position differences within 3–4 cm demonstrate acceptable model scaling
• Segment angle differences under 5° (especially in sagittal/frontal planes) indicate strong alignment
• Joint angle differences under 5° are in line with literature standards, especially for joint angles with stronger signals, like knee flexion

If you're comparing systems, make sure vendors are transparent about how they validate their accuracy, and whether their results hold across real-world conditions, not just in controlled labs.

Don’t Overlook Repeatability

Even a perfectly accurate system loses value if it can’t deliver consistent results.

Repeatability refers to how similar the output is when the same person performs the same movement, under the same conditions, multiple times. It matters because in research and applied sport, inconsistent data can kill your analysis.

STRN emphasizes that systems should produce repeatable data with low inter-trial variability. That means:

• Repeat measurements for the same subject should yield nearly identical joint angles
• Differences should fall within the natural variability of human movement (typically ≤2–3°)
  

Studies using Theia3D report inter-trial variability similar to or better than marker-based systems — particularly in walking, running, and upper-limb tasks.

Reproducibility: The Multi-Site Question

Here’s where things get real.

Reproducibility asks whether your system can still perform under varying conditions, such as when:

• Different users are collecting data
• Subjects wear different clothing
• Lighting, camera setups, or lab spaces vary
  

If you’re running a multi-site study or want to compare results over time, reproducibility is non-negotiable.

Unlike marker-based systems, where small differences in marker placement can throw off joint kinematics, markerless approaches like Theia3D don’t rely on the operator. That means reproducibility across sessions and sites tends to be much tighter.

Look for systems that report inter-session variability (with attire changes) and test across different environments. Bonus points if this data is published in independent peer-reviewed studies.

Technical Specs: What’s “Good Enough”?

Specs can feel like a checklist, but they actually tell you what kind of data quality you can realistically expect.

According to STRN, vendors should be transparent about their tech’s capacities. For markerless systems:

• Minimum video resolution (e.g., 500+ pixels in height per subject)
• Camera count, frame rate, and field of view requirements
• Recommended lighting conditions
• Required calibration precision
  

These affect everything from whether a system can handle fast sports movements to how portable it is across spaces. Ask vendors: What happens if the camera angle shifts? If lighting changes? If we use a different camera model?

Theia3D, for example, supports a variety of camera systems as long as resolution, calibration, synchronization, and visibility requirements are met. But not every system is this flexible.

Summary: What This Pillar Tells You

When evaluating any motion capture system, start with this question:

Can I trust the data — across users, environments, and time?

If the answer is anything less than “yes,” it’s a red flag. STRN’s first pillar gives you a framework to vet those claims, and this post gives you the key questions to ask.

If you're ready to dive deeper or want a complete checklist to compare vendors, download our full Buyer’s Guide below.
‍