WEEK 5

Why thresholds matter

A system needs a definition of “significant”

Handwriting is continuous. Pressure shifts gradually. Speed fluctuates subtly. Direction curves fluidly. But for a system to respond, it must decide when behaviour becomes significant.

• When does pressure become emphasis?
• When does slowdown become hesitation?
• When does turning become reorientation?

At the same time that I defined these behavioural signals, I also defined their geometric responses. Detection and reconstruction were not separate stages. They were developed together. I finalised three behavioural signals and their corresponding forms:

Rectangles form the stroke backbone. All shapes share a consistent base width proportional to glyph height. This constraint keeps the system unified even when behaviour varies. As the system detects, it constructs. As behaviour is segmented, geometry appears. This is where the project shifts from observing gesture to rebuilding it through rule-based form.

Manual test — letter “g”

Proving the concept before coding it

Before coding the logic, I tested the idea manually. I asked three friends and myself to write the letter “g” using a Wacom tablet. I recorded the traces and manually analysed them in Photoshop. Using my own logic, I placed:

The idea worked conceptually before it worked computationally. This was the first time the glyph idea felt like it had potential. This manual mapping confirmed two things:

• The three behavioural signals were visually identifiable.
• Different writers produced different behavioural patterns, even when writing the same letter.

Introducing thresholds

Discretising continuous behaviour

The next step was to make the system detect these behaviours automatically. I implemented thresholds:

• Pressure peaks
  relative to the writer’s
  baseline pressure
• Hesitation
  relative to mean
  writing speed
• Reorientation
  based on angle
  change threshold

Breakthrough + friction

Coherence mattered more than perfect accuracy

The first threshold-based outputs were slightly chaotic. Some detections were off. Some shapes overlapped awkwardly. It took several adjustments. But when I tested it on my own handwriting and saw the glyph respond meaningfully, something shifted.

Realisation: The form was not perfectly accurate, but it was coherent. Mis-detections did not destroy the form. They altered it. That alteration felt interpretative rather than broken.

This was the breakthrough moment. Week 5 was a breakthrough because I finally formalised the system I wanted to build: a rule-based behavioural glyph engine.

Design stance

Choosing interpretation over optimisation

I began questioning whether the system needed to be accurate. Accuracy would mean perfectly replicating pressure, timing, and motion. That would shift the project toward engineering optimisation.

Instead, I chose interpretation. The algorithm is authored. It follows my rule set. It translates behaviour into a geometric language. If behavioural signals are legible but the original letter becomes less readable, that is acceptable. Behaviour is prioritised over letterform.

The system is not reproducing handwriting. It is reconstructing behavioural traces.

Behavioural mirror

What becomes visible through the glyph

When users test the system, they begin to notice:

• Where they press harder
• Where they slow down
• How their curves differ from others
• Where they pivot sharply

The glyph becomes a behavioural mirror. I also want users to compare outputs between writers. Writing the same letter should produce different glyphs. The variation is the point.