Tangible Interaction / Non-Visual Interface / Hardware

Braille Wordie

Duration

Oct - Dec 2023 · 2 months

Type

Group Work · UC Berkeley

Skills

Tangible Interaction, Circuit Prototyping, 3D Modeling

Tools

Rhino, Arduino, 3D Printing

No screen, no cursor. Braille Wordie is a tangible interface for blind children: place a block by hand, hear the result, and practice braille without a teacher in the room.

A two-month group project at UC Berkeley: 3D-printed letter blocks, snap-together boards, three buttons and a speaker, driven by an Arduino.

Team demo: from opening the box to spelling a word.

A tactile skill can't be taught on a screen

Braille is learned through the fingers, with constant correction. An app can't put dots under a fingertip and teachers are scarce, so the material itself had to give feedback.

1
17 US states report a shortage of Visual Impairment teachers. At Kentucky School for the Blind, the student/teacher ratio reached 61:1.
2
Virtual classes made it worse: touch doesn't travel over video.
3
Traditional braille materials give no real-time feedback, so students can't self-correct.
4
Existing teaching aids allow one fixed exercise each.

Why blocks? An app has no dots. A refreshable braille display is expensive and shows one line at a time. Embossed cards stay silent. Only blocks gave touch, sound and rearrangement at once.

My part in it

A team of three, with Charon Kuo and Tuki Han. I did the Rhino modeling, 3D printing, wooden case and assembly, and designed the three-button control, audio menu and lesson flow. I also shot and voiced the demo video and ran testing at the showcase. For the planned AI learning assistant, I set up and trained an LLM on the OpenAI Platform (Assistants and fine-tuning) with medical content and a fictional children's learning plan, by uploading source files, writing its instructions and preparing Q&A data. It was not integrated into the prototype.

Audio is the display, blocks are the input

Without a screen, three channels are left:

1
Input: drop letter blocks into board slots and snap boards together.
2
Output: instructions, pronunciations, praise and corrections, all spoken.
3
Navigation: back, select and forward buttons drive an audio menu.
Two boards connected to spell REDO, left hand on the button cluster while the right hand rests on the blocks

Designing for hands, not eyes

Every part had to make sense to a fingertip:

1
Two triangular arrows flank a round select button, so direction is clear by touch.
2
Board wells are textured and block faces smooth, so a hand can tell if a slot is filled.
3
Each block has raised braille dots and an embossed letter, so a blind child and a sighted parent can share it.
4
A light sensor starts the spoken welcome when the lid opens, so no adult is needed to set it up.
Close-up of a fingertip on the three-button cluster, triangular arrows and a round select button
Fingertips reading the raised dots on a letter block
A hand gripping a letter block above the board's textured well

Snapping boards together is the grammar

Each board holds two blocks, and boards connect side by side. Spell RE on one, snap on DO, and the device plays the combined word. A prefix becomes a board you can attach. We also scoped an ED board for past tense and chained boards for sentences.

Two hands bringing a second board toward the first to spell REDO
Boards spelling UNDO after swapping the prefix blocks, RE blocks set aside
Scoped scenario diagram: snapping an ED board onto PLAY to form the past tense
Scoped scenario diagram: chaining boards into the sentence I AM AMY

Sized to hands, built around a microcontroller

1
Blocks are 30 × 50 × 12 mm with Ø8 mm dots, sized for children's 52-62 mm hands and 30 mm maximum grip.
2
Fully 3D-printed, modeled in Rhino, Arduino-driven, and wired by the team.
3
Boards, blocks and electronics all pack into one wooden case.

An RFID tag in each block tells the board which letter was placed. The weak point was the wiring: audio and data ran over wires, which made the experience slower and less intuitive.

3D-printed letter blocks beside the dimension drawing: 30 by 50 by 12 millimeters with 8 millimeter dots

Three sessions, one loop

Tutorial introduces the object, Basic spells a word and plays its sound and meaning, and Advanced changes words with prefixes. Every step is the same loop: act, listen, correct, go again.

Tutorial: opening the box triggers the welcome message
Basic session: the device introduces the lesson by voice
Basic session: audio task instruction, find the letter R block
Basic session: spelling RE with blocks, confirmed by voice
Basic session: hearing the definition of the word just spelled
Advanced session: following the audio instruction to switch the prefix

Did it hold up in real hands?

We tested the prototype at the Jacobs Winter Design Showcase 2023. Sound that played the moment a block was placed confused people, so the flow became button-led with a spoken menu. Critics also pointed out that our braille cells were too large for one fingertip, so the next version should use standard-size cells.

Demoing the working prototype

What screen-less interaction taught me

1
State must be announced. Without a screen, what the system doesn't say doesn't exist.
2
Shape, texture and size have to explain the system. There is no tooltip.
3
Recovery is physical. A wrong block must be as easy to lift out as it was to place.

Robots, wearables and appliances face the same question: how can a body read a machine's state and correct its mistakes?

What's next

Scoped but not built: past tense and sentence writing on the same boards, and an AI text-to-audio assistant that turns any book into braille practice.

Concept render of the production design, color-coded blocks on two connected boards
Where it fell short
Braille cells were too large for real-book practice, and we only had showcase feedback, not a structured study with blind or low-vision learners.
Thanks
Charon Kuo and Tuki Han, who built Braille Wordie with me.