Decoding Intent Through Mechanical Muscle Deformation

19 concepts11 papers1 proposal

When we move a limb, our brains send electrical impulses to our muscles, prompting them to contract. For decades, human-machine interfaces have relied on surface electromyography (sEMG) to listen to these electrical commands right at the skin's surface. While elegant in theory, sEMG is notoriously fragile in practice. The signals are tiny—measured in microvolts—and easily corrupted by sweat, skin impedance changes, and cross-talk from neighboring muscles. If a sensor shifts by even a few millimeters, the system's calibration often breaks entirely.

A growing body of research shifts the focus from the electrical command to its physical execution. When a muscle contracts, its fibers shorten and thicken, resulting in a macroscopic, volumetric change in shape. By mapping this physical deformation, researchers are building interfaces that are inherently stable, highly intuitive, and remarkably resilient to environmental noise.

The Modalities of Physical Mapping

Mechanical muscle sensing captures the structural geometry of movement through several distinct modalities, each offering a unique window into tissue deformation.

Surface Pressure and Topographic Force Mapping

Force Myography (FMG) and Tactile Myography (TMG) utilize arrays of pressure-sensitive resistors or high-resolution tactile cells worn around a limb. As muscles expand during contraction, they press outward against the sensor band.

Cho et al. demonstrated that this volumetric muscle deformation provides a high-fidelity signal for intent recognition, achieving nearly 90% accuracy in classifying complex grip patterns for transradial amputees. Because it tracks physical displacement rather than electrical activity, the signal remains stable even when physiological factors like sweat would otherwise degrade an sEMG system. To enhance precision, Jaquier et al. combined TMG with traditional sEMG, demonstrating that capturing physical muscle bulging provides distinct, complementary data that significantly improves proportional hand and wrist activity detection.

Acoustic and Vibrational Sensing

Mechanomyography (MMG) measures the low-frequency mechanical vibrations generated by the physical oscillation of muscle fibers during contraction. This approach acts almost like a microphone listening to the internal machinery of the limb.

Gardner et al. integrated MMG into a multimodal sensor suite for upper-limb prostheses, highlighting its natural immunity to skin conditions and motion artifacts. This immunity is particularly vital in closed-loop Functional Electrical Stimulation (FES) systems. In these setups, external electrical currents are used to stimulate paralyzed muscles, completely blinding traditional sEMG sensors with massive electrical artifacts. Woods et al. bypassed this limitation entirely by using MMG to monitor muscle state and fatigue during active stimulation, proving that mechanical vibration sensing can provide clean feedback in environments with extreme electromagnetic noise.

Morphological and Deep Tissue Imaging

Instead of measuring surface pressure or vibration, some systems look directly at the changing shape of the limb. Bai et al. noted that deformation-based sensing—using arrays of distance sensors to track forearm shape changes—offers a robust alternative to computer vision, which suffers from camera occlusion, and sEMG, which suffers from signal noise.

Taking this concept deeper, ultrasound imaging allows systems to visualize internal structural shifts in real time. Bimbraw et al. utilized forearm ultrasound imaging to capture fine-grained muscle activity, successfully predicting metacarpophalangeal joint angles and hand configurations without impeding natural movement.

[User Intent] ──> [Muscle Contraction] ──> [Volumetric Deformation]
                                                    │
                   ┌────────────────────────────────┼──────────────────────────────┐
                   ▼                                ▼                              ▼
         [Surface Pressure]                [Internal Imaging]             [Acoustic Vibration]
         • Force Myography (FMG)           • Ultrasound Imaging           • Mechanomyography (MMG)
         • Tactile Myography (TMG)         • Distance/Strain Sensors      • Microphones/Accelerometers

Why Mechanical Signals are Inherently Robust

To understand why mechanical sensing works so reliably, it helps to contrast the underlying physics of the signals. An electrical action potential is a transient, chemical event that must pass through highly resistive layers of fat and skin. A mechanical signal, by contrast, is a structural change in the shape of the tissue itself. This fundamental difference yields several distinct advantages:

  • Placement Insensitivity: Because an electrical signal is highly localized to the motor unit, moving an sEMG electrode slightly can cause the signal to vanish. Mechanical volume changes, however, are distributed across the entire muscle belly. Marquardt et al. demonstrated that FMG-based torque estimation in human knee and ankle joints remains highly accurate across varying movement speeds, proving less sensitive to precise sensor alignment and skin conditions than bioelectric alternatives.
  • Hardware Optimization: Because the physical signal is so pronounced, the underlying hardware can be simplified. Xu et al. developed a graph-based framework to optimize FMG armband design, proving that just 4 strategically placed force sensors could achieve over 93% accuracy in arm movement recognition—nearly matching the performance of a dense 16-sensor array. Even ultra-low-profile form factors, such as the graphene-nanocomposite "Motion Tape" evaluated by Lee et al., can successfully capture these macroscopic muscle bulges to track complex joint kinematics.
  • Zero-Shot Generalization: In traditional bioelectric interfaces, differences in skin impedance and anatomy mean a model trained on one person rarely works on another without extensive recalibration. However, the geometric relationship between muscle deformation and joint kinematics is highly consistent across human anatomy. Xingchen Yang leveraged this property to build a "plug-and-play" ultrasound armband for virtual reality tracking. By treating multi-channel ultrasound signals as morphological images, the system achieved zero-shot generalization across different users and sessions without requiring individual calibration, successfully overcoming the "domain shift" hurdle that has long restricted wearable muscle-sensing interfaces.

By anchoring human-machine interaction in the physical reality of muscle deformation rather than the volatile environment of surface biopotentials, these diverse mechanical approaches offer a stable, plug-and-play pathway toward truly intuitive robotic control and motion tracking.

Key Foundational Work

Force Myography to Control Robotic Upper Extremity Prostheses: A Feasibility Study (2016) — contributed 2 concept elements to this direction.
Proceedings of the first workshop on Peripheral Machine Interfaces: going beyond traditional surface electromyography (2014) — contributed 2 concept elements to this direction.
History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 11 papers
  2. Version 1 · 2026-10-04 · 11 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_force_myography_fef1dc10,
  title = {force myography},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/force-myography/rdr_fef1dc101f9947f4aa334ca6d0daab75},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 2, frozen text):

@misc{lacuna_force_myography_fef1dc10_v2,
  title = {force myography},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/force-myography/rdr_fef1dc101f9947f4aa334ca6d0daab75/v/2},
  note = {Version 2, frozen 2026-10-04. Accessed 2026-11-21},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake