Motion: turning movement into strokes

Pressure, tilt, velocity, rhythm, each is a thread. Motion is the cloth they're woven into: the complete, time-ordered path of the pen tip across the page. Recover that, and you've recovered the handwriting itself.

The master signal

Every other signal is a view onto motion. Velocity is how fast the motion is; tilt is the orientation the motion happens in; pressure is the force behind it; rhythm is its timing. Reconstructing the full trajectory, where the tip was at every instant, is therefore the central task, the one that makes a faithful digital record of writing on plain paper possible.

From inertial data to a path

In principle the recipe is simple: take the pen's orientation (so you know which way "forward" is), rotate the measured acceleration into page coordinates, subtract gravity, then integrate twice, once to get velocity, again to get position. In practice this double integration amplifies noise badly; raw physics alone drifts off the page within seconds.

So modern systems combine physics with learning. The dual-IMU geometry and sensor fusion supply a clean orientation; zero-velocity resets at pen-lifts bound the drift; and a trained model maps the inertial stream to tip displacement, correcting the systematic errors that hand-built equations miss. The result is a trajectory that holds together across many strokes, not just one.

Stitching strokes and lifts together

Real writing isn't one continuous line, it's strokes separated by lifts and slides. Motion reconstruction works hand in hand with intent detection (is the pen writing or hovering?) to decide which segments are ink and which are travel. Get that segmentation right and the page reads as crisp, separate strokes; get it wrong and phantom lines connect letters that were never joined.

Rendering Indic conjuncts faithfully

Nowhere is a faithful path more important than in Indic scripts. A single character can stack a consonant, a vowel sign and a conjunct into one dense glyph: think of the layered aksharas of Kannada and Telugu, the flowing curves of Malayalam, the loops of Tamil, or the conjunct clusters hanging beneath Devanagari's head-line in Hindi. The order and path of strokes within these glyphs carry meaning and reveal skill.

If reconstruction blurs which stroke came first or where a loop closed, the rendered akshara can read as a different character entirely. Capturing motion precisely is what lets Vahini represent these scripts as they were written, and score how cleanly each complex character was formed.