The Vahini research journal
Questions worth
working on.
Handwriting, sensing and the experiments behind them. Read a practical explanation, inspect the diagram, and follow the source.
31 notes · diagrams, methods and sources
Product explainer
The 20 factors: measurements, proxies and practice
Which parts of handwriting can a page image actually tell us about?
2 min read · Malli Kosuri →
Product explainer
From a handwriting photo to useful practice feedback
How does a page image become handwriting feedback?
2 min read · Malli Kosuri →
Research note
Handwriting screening research: evidence and limits
What would it take for a handwriting model to support screening responsibly?
2 min read · Vishnu Kosuri →
Research note
Studying handwriting support with autistic learners
How can writing research respect the needs of individual learners?
2 min read · Malli Kosuri →
Founder note
A research-led pitch: connect claims to evidence
How can a deep-tech pitch distinguish a useful prototype from a finished product?
2 min read · Vishnu Kosuri →
Engineering note
One IMU or two: what the second view adds
What information does a second motion sensor add to a rigid pen?
2 min read · Malli Kosuri →
Engineering note
Reading force: calibration before interpretation
How does a force-sensitive element become a useful tip-force signal?
2 min read · Vishnu Kosuri →
Engineering note
Estimating pen tilt while the hand is moving
When can an accelerometer describe tilt, and when does writing motion interfere?
2 min read · Malli Kosuri →
Engineering note
Why pen speed is difficult to recover from acceleration
Why does a small acceleration error become a large writing error?
2 min read · Vishnu Kosuri →
Engineering note
Turning motion samples into separate writing strokes
How can a continuous sensor stream become a sequence of strokes?
2 min read · Malli Kosuri →
Engineering note
Measuring writing rhythm without judging the writer
What can the timing of strokes tell us about a writing task?
2 min read · Vishnu Kosuri →
Engineering note
Pen down or pen up: detecting contact reliably
How do we avoid turning a hovering movement into a written stroke?
2 min read · Malli Kosuri →
Engineering note
What a compass can add to pen orientation
When should an orientation estimator trust a magnetic reading?
2 min read · Vishnu Kosuri →
Engineering note
Streaming motion over BLE without losing the timeline
How do we know whether a received stream still represents the original recording?
2 min read · Malli Kosuri →
Research note
Reconstructing handwriting: define the target first
Are we trying to reproduce the ink path, recognise text, or both?
2 min read · Vishnu Kosuri →
Research note
What ordinary-paper capture needs to prove
How do we evaluate a pen without quietly relying on a special writing surface?
2 min read · Malli Kosuri →
Engineering note
104, 208 or 416 Hz: choosing a sampling rate
What does doubling the sampling rate buy, and what does it cost?
2 min read · Malli Kosuri →
Paper note
The image and the strokes: two views of handwriting
Why might a model need both the final image and the stroke sequence?
2 min read · Vishnu Kosuri →
Paper note
Handwriting health signals: evidence and limits
When does an association in a study become useful beyond that study?
2 min read · Malli Kosuri →
Research note
Removing a sensor: an experiment, not a promise
When can a simpler design preserve the performance we need?
2 min read · Malli Kosuri →
Research note
Testing recognition on writers the model has never seen
Can the recogniser read a new writer, or has it memorised familiar hands?
2 min read · Vishnu Kosuri →
Research note
More data is useful only if evaluation stays honest
How do we tell whether a larger dataset actually improves the model?
2 min read · Malli Kosuri →
Engineering note
The sensor moves differently from the pen tip
Why is tracking a sensor in the barrel not enough to track the ink?
2 min read · Malli Kosuri →
Engineering note
Calibrating a compass near real desks
What can calibration correct, and what changes when the environment changes?
2 min read · Malli Kosuri →
Engineering note
Using motion events to wake a sleeping processor
Which events should be recognised before the main processor wakes?
2 min read · Malli Kosuri →
Engineering note
Why tip contact needs its own measurement
What ambiguity does a tip-force signal remove from a motion recording?
2 min read · Malli Kosuri →
Engineering note
Measuring the power cost of every pen state
Which state dominates battery use during a real day?
2 min read · Malli Kosuri →
Engineering note
Choosing Bluetooth for a sensor stream
Why choose Bluetooth Low Energy for a pen, and what still needs testing?
2 min read · Malli Kosuri →
Engineering note
Battery runtime and charging: what to measure
What evidence belongs behind a battery-life claim?
2 min read · Malli Kosuri →
Engineering note
Why the refill belongs in the measurement model
How can changing an ordinary refill change a sensing experiment?
2 min read · Malli Kosuri →
Engineering note
A sensor pen is also a real-time data system
What must the software do when samples arrive faster than they can be sent?
2 min read · Malli Kosuri →No matching notes. Try another word or topic.
Read it. Try one thing.
The live analyser works with a photo of your handwriting. The sensor pen and motion-recognition work have their own development path.
Check your handwriting →Explore Project Akshara →