Movement practitioners considering wearable sensing usually ask which device to buy. The more useful question is which signal type the device produces, because that determines what can be measured at all. This guide sets out four signal types available in consumer and prosumer hardware — inertial, positional, force/pressure and electromyographic — explains what each can and cannot capture, and gives a decision path for choosing among them. It is written for dancers, somatic educators and movement researchers with no engineering background, and reflects hardware and research available as of July 2026.

Why signal type matters more than device brand

A sensor can only report what its physical principle allows it to detect. No amount of software processing recovers information that was never in the signal. This is not a limitation of current products; it is a property of the measurement.

This point was quantified in robotics research in July 2026, when a benchmark testing humanoid robots under real applied human forces found that vision-based control systems failed systematically, because camera images do not encode contact force, friction or the onset of slipping — the analysis is covered in the Lab's explainer on why a robot can see your hand but not feel it. The practical lesson generalises: before choosing hardware, decide which phenomenon you are trying to observe, then choose a signal type that physically carries it.

The four signal types

1. Inertial (IMU) — acceleration and orientation

What it is. An inertial measurement unit combines an accelerometer and gyroscope, reporting how fast a body segment is accelerating and how it is oriented. Present in every smartphone and smartwatch.

Captures well: movement dynamics over time — speed, acceleration and deceleration profiles, rhythmic structure, sudden versus sustained qualities, orientation changes. Because Laban's Time effort factor is largely a matter of acceleration profile, IMUs capture that dimension reasonably directly.

Cannot capture: absolute position in a room without additional reference; muscular effort; contact with another body or surface. An IMU cannot distinguish a movement performed with high muscular co-contraction from a visually identical one performed freely, because both produce similar accelerations.

Practical note: IMU drift accumulates — orientation estimates degrade over minutes without correction. Research published in July 2026 (arXiv:2607.09780) demonstrated full-body motion reconstruction from arbitrary subsets of ordinary consumer devices — phone, watch, glasses, insoles — with a model designed to handle whichever sensors happen to be present. This is the most practically relevant direction for practitioners: the hardware you already own may be sufficient for useful movement reconstruction.

2. Positional (camera, markerless capture) — where the body is

What it is. Camera-based systems estimating joint positions in space, increasingly without markers.

Captures well: spatial pathways, body shape and configuration, relationships between body parts, floor patterns. Multi-camera markerless systems now approach marker-based accuracy, including for multiple interacting people, as covered in the Lab's account of the end of the motion capture suit.

Cannot capture: force, effort, contact dynamics, or anything occluded. Occlusion is the practical constraint — floor work, partnering and any movement where the body folds onto itself will produce gaps.

Practical note: a single camera gives useful 2D analysis and rough 3D estimation; reliable 3D needs multiple synchronised views. Recent prediction models tolerate partial and occluded input better than earlier systems (arXiv:2607.10984), which reduces but does not eliminate the problem.

3. Force and pressure — what the body pushes against

What it is. Force plates, pressure mats and pressure-sensing insoles measuring force exchanged with the ground or another surface.

Captures well: weight distribution and shifts, ground reaction force, balance and postural sway, the timing and magnitude of impact on landing. For any practice concerned with grounding, weight transfer or fall and recovery, this is the signal type that carries the phenomenon.

Cannot capture: movement away from the measured surface; internal muscular organisation.

Practical note: force plates are laboratory equipment; pressure insoles are the accessible consumer entry point and are adequate for weight-shift and balance work.

4. Electromyography (EMG) — muscle activation

What it is. Surface electrodes on the skin detecting electrical activity as muscles activate.

Captures well: which muscles are working, how hard, and — importantly — when, including activation that precedes visible movement by roughly 50–200 milliseconds. EMG also allows measurement of co-contraction, the simultaneous activation of opposing muscles that distinguishes held, controlled movement from released, flowing movement. This is the signal type that reaches the layer where movement quality is physically constituted, an argument developed in the Lab's analysis of muscle redundancy and movement quality.

Cannot capture: position or spatial pathway; deep muscles beneath surface layers; and it does not directly capture felt experience — it measures the electrical correlate of muscular effort, not the sensation itself.

Practical note: EMG is the most demanding option. Signal quality depends on electrode placement, skin preparation and movement artefact; readings vary substantially between individuals and between sessions, so per-session calibration is necessary. Research-grade systems are expensive; consumer-grade options sample more slowly and are noisier. Do not expect plug-and-play results.

A decision path

Work from the question, not the device.

  • "How does this movement travel through space?" → positional (camera). Multi-camera if the movement involves floor work or partnering.
  • "What is the dynamic quality — sudden, sustained, rhythmic?" → inertial. A phone and watch may suffice.
  • "Where is the weight, and how does it transfer?" → force/pressure. Insoles are the accessible option.
  • "How much effort, and is the movement held or released?" → EMG. Accept the setup burden, or reconsider whether the question can be answered another way.
  • "What is the mover experiencing?" → no sensor answers this. Use the practitioner's own report and trained observation. Sensors provide correlates, not experience.

Limitations to hold onto

Three constraints are worth stating plainly, because vendor material rarely does.

Sensors measure correlates, not experience. EMG measures electrical activity accompanying muscular effort. It does not measure how effortful a movement felt. The correlation is real and useful; the identity is not.

More sensors is not better. Each added stream increases setup time, synchronisation difficulty and analysis burden. A well-chosen single signal type usually beats a poorly integrated multi-modal rig.

Data does not replace trained perception. A practitioner's observation remains the instrument that detects qualities no current sensor captures — initiation, whole-body connectivity, intention legibility. The Lab's framework for assessing generated movement sets out those dimensions in detail. Sensing extends what can be recorded; it does not extend what can be noticed.

References

Curreli, C., Hofherr, F., Muhle, D., Saroha, A., Marin, R., & Cremers, D. (2026). EquiFusion: Kinematics-agnostic human motion prediction via equivariant latent diffusion. arXiv:2607.10984. https://arxiv.org/abs/2607.10984

Towards real-world wearable motion reconstruction. (2026). arXiv:2607.09780. https://arxiv.org/abs/2607.09780

Yu, C., et al. (2026). ThorArena: Benchmarking humanoid physical interaction with human motion-force demonstrations. arXiv:2607.06052. https://arxiv.org/abs/2607.06052