All posts Technology

One IMU or two? Why dual-IMU sensing is more accurate

MKMalli KosuriJun 20266 min read
In short

A single IMU can't separate a true tilt of the pen from the pen sliding across the page, both look like the same blend of acceleration and rotation. Vahini places two IMUs at different points along the barrel, so the difference between their readings reveals the pen's real rotation and the position of the tip independently of how it's moving.

That extra constraint sharpens tilt, velocity and tip-path estimates, and feeds cleaner numbers into all twenty handwriting factors, especially for the dense, curved letterforms of Indic scripts.

The short version. The full post has the detail and the why.

In this post

  1. The ambiguity one sensor can't solve
  2. Two IMUs, two vantage points
  3. From two readings to one clean orientation
  4. What it sharpens downstream

An inertial measurement unit (IMU) feels acceleration and rotation. It's a remarkable little chip, but on its own, inside a moving pen, it faces a problem it genuinely cannot solve.

The ambiguity one sensor can't solve

An accelerometer measures all acceleration at once, the pen speeding up, and the constant tug of gravity. When the pen rotates, the gravity component shifts across its axes; when the pen slides, real motion shows up on the same axes. From a single point, a tilt and a slide can produce nearly identical readings. The gyroscope helps by measuring rotation rate directly, but it drifts, and one gyro still can't tell you where along the pen a motion originated.

The result, with one IMU, is that the pen is always making an educated guess about whether you tipped it or moved it. For casual motion tracking that's fine. For measuring the millimetre-scale geometry of handwriting, the guesswork shows up as noise in every factor.

Two IMUs, two vantage points

Put a second IMU further down the barrel and the picture changes completely. When a rigid pen rotates, the two units, separated by a known distance, experience different accelerations. That difference is a direct, geometric fingerprint of rotation about the tip. When the pen merely slides, both units see the same motion. Comparing the two readings cleanly separates "the pen turned" from "the pen moved."

This is the same principle that lets your inner ear sense head rotation, and the reason aircraft and robots use distributed inertial sensing. The known lever arm between the two IMUs turns an ambiguous signal into two equations with one answer.

IMU 1 IMU 2 rotation:readings differ slide:readings match
Two IMUs separated by a known distance: rotation produces a measurable difference between them, while a pure slide moves both identically. That contrast is what resolves the single-sensor ambiguity.

From two readings to one clean orientation

The two streams are fused, typically with a Kalman-style filter, into a single best estimate of the pen's orientation and the motion of its tip. Each sensor checks the other: the gyroscopes constrain fast rotation, the accelerometers anchor the slow drift, and the spatial separation pins down the geometry. The pen ends up with a steadier, less noisy sense of how it is held and where the tip is going than any single unit could provide.

What it sharpens downstream

Cleaner orientation isn't an end in itself, it's the foundation the handwriting factors stand on. Slant and vertical alignment depend on knowing the true tilt. Velocity and rhythm depend on a tip path that isn't polluted by rotation artefacts. Stroke reconstruction depends on all of it.

This matters most for Indic scripts, Malayalam, Telugu, Kannada, Tamil and Devanagari for Hindi, where letters are built from tight curves, loops and stacked conjuncts. A small orientation error that a Latin print sample would shrug off can blur the difference between two similar aksharas. Dual-IMU sensing buys back that precision.

Key takeaways

  • One IMU can't reliably separate tilting the pen from sliding it.
  • Two IMUs a known distance apart turn that ambiguity into solvable geometry.
  • Fusing both yields steadier orientation and a cleaner tip path.
  • The precision gain matters most for the curved, stacked letterforms of Indic scripts.

It's the reason Vahini's pen carries two inertial units rather than one. The second sensor isn't redundancy, it's the piece that makes the measurement trustworthy.

Read next