How Do Data Gloves Turn Hand Motion Into Data?

Bend a finger. How can a data collection glove describe how far it bends? EP04 uses general IMU principles to explain the steps between motion signals and a meaningful hand-motion record.

An IMU, or inertial measurement unit, turns with the finger and senses rotation and acceleration. Algorithms use these signals to estimate orientation. Sensor orientation is not the angle of every joint. Each sensor first needs a relationship to the finger segment it follows. Known poses establish a calibration reference. Orientation comparisons and a joint model can then help estimate bending.

Calibration does not guarantee accuracy. Checking other reference movements can reveal a difference between calculated and real motion. That difference may affect how a record describes the fingertip approaching a cup. Finding the cause, adjusting, and checking again helps establish what the record means.

Joint angles describe bending. Fingertip pose describes position and orientation. A model, including finger lengths, can help estimate pose; what can be recovered depends on the measurements and model. Readings alone cannot fill in every movement of the hand.

Position also needs a reference. Move the whole hand while keeping its shape fixed: a fingertip may stay still relative to the wrist but move relative to the table. IMU orientation estimates are not reliable global position. Describing the whole hand in external space requires an appropriate tracking system.

Saving measurements over time describes opening and closing. Suitable whole-hand tracking can add movement through space. Hand-motion records do not automatically contain force measurements.

The video uses animated, general-principle diagrams rather than a specific PHAIN product or performance demonstration. Next in the series, we explore what head and wrist cameras can add to a task record.