Skip to main content
Many assistive technologies address one part of the blind student’s writing problem. Some provide a tactile input method. Others output audio. A few connect to a teacher’s screen. What makes BrailleBridge different is that no single existing product delivers the specific combination it offers: a physical six-dot Braille keyboard, bilingual English and Bangla translation, Bluetooth HID output to the student’s own device, and a professional web-based Teacher Software — all within one system that costs under $15 to build and can be repaired with a screwdriver and a few dollars of standard parts.

What makes BrailleBridge different

Six-dot physical Braille input

Real mechanical Braille chording on tactile Outemu Red switches — not a touchscreen, not a virtual keyboard. The student writes Braille the way it was meant to be written: by feel, with both hands, using the muscle memory they already have. The switches are rated for approximately 50 million presses and provide a consistent, linear action that blind users can rely on.

English + Bangla from the same hardware

One device, two complete dictionaries. A Shift-based modifier system provides access to Bangla vowel signs (কার/matra), digits 0–9, punctuation, and mathematical operators — all without any hardware change. Language switching is a deliberate chord hold (dots 4+5+6 for 500 ms), making accidental switching practically impossible.

Bluetooth HID keyboard output

The ESP32 registers as a standard Bluetooth HID keyboard — the same profile used by every wireless keyboard on the market. The student’s phone, tablet, or computer pairs with it like any other keyboard: no proprietary receiver, no custom driver, no app required. Any device with a screen reader can immediately speak the output.

Web-based Teacher Software via USB Serial

A professional HTML/CSS/JS application reads the ESP32’s USB Serial output through the browser’s Web Serial API. It requires no installation, no server, and no database. The teacher opens a single HTML file in Chrome or Edge, connects the USB cable, and sees every student’s translated text appear in real time with timestamps and language indicators.

ESP32 architecture at ~$5

The ESP32 Dev Module — roughly $5 per unit — provides a dual-core processor, 520 KB SRAM, 4 MB flash, integrated Bluetooth, and a USB-UART bridge in a single package. Outemu Red switches cost approximately $0.38 each. The entire bill of materials totals around $12 per device, making school-wide deployment realistic on limited budgets.

Custom 3D-printed enclosure

Designed parametrically in Fusion 360 and printed in PLA on any FDM printer (0.2 mm layers, ~30 g of filament per device). MX-compatible switch cutouts mean the Outemu switches snap in without glue. Rounded corners, raised tactile keycap markings, LED light channels, and internal space for a future Li-ion battery are all part of the design. The CAD files and STL models are open source — any school or makerspace can print a replacement.

Designed specifically for classrooms and examinations

Every design decision — silent input, dual output, timestamped records, exam mode, multi-student monitoring, language indicators, failsafe character storage — was made with the specific context of a classroom examination in mind. BrailleBridge is not a general-purpose accessibility device repurposed for education; it was built for education from the ground up.

Comparison with existing solutions

No existing solution delivers the same combination of features at BrailleBridge’s price point. The table below shows how it compares to the alternatives a school might actually consider.

Engineering innovations

The engineering choices behind BrailleBridge are as important as the features they enable. Each innovation below solves a specific problem that would otherwise make the system unreliable, expensive, or difficult to maintain.
Human fingers pressing a chord do not land at the exact same microsecond. Without a collection window, a single intentional chord (e.g., dots 1 and 4 for the letter C) could be misread as two separate single-dot presses. The 180 ms window starts on the first key press and ORs every additional key pressed within that window into the same chord value. When the window expires, the complete chord is processed. This single decision makes reliable chording possible on affordable, non-synchronised mechanical switches.
Each of the six Braille dots is assigned a unique bit position in an 8-bit integer (dot 1 = bit 0, dot 2 = bit 1, … dot 6 = bit 5). When a key is pressed (GPIO reads LOW), its corresponding bit is set with a bitwise OR. The resulting 6-bit value is the chord — a single integer that uniquely identifies any of the 63 possible chord combinations. Lookups are constant-time comparisons of a single integer against an array, making the translation fast enough that latency is imperceptible even on a large dictionary.
Rather than hard-coding translations in switch-case blocks, the firmware stores all mappings in a structured array of BrailleEntry records, each containing a chord value and its English and Bangla outputs. A lookup() function performs a linear search and returns the correct string. The critical implication: adding a new character, a new language, or a new symbol set requires only appending entries to the array — no control-flow changes, no recompilation of core logic, no firmware restructuring.
Every translated character is sent through two completely independent channels in the same firmware call: over Bluetooth HID to the student’s paired device, and over USB Serial at 115200 baud to the Teacher Software. The two channels are decoupled — if Bluetooth drops, the serial record is unaffected; if the USB cable is unplugged, the student’s audio feedback continues. This independence is critical for examination integrity.
Language selection is controlled by a single isBangla boolean variable in the firmware, toggled by a deliberate 500 ms hold of dots 4+5+6. This chord was chosen because no standard letter, number, or symbol chord uses exactly dots 4+5+6, making accidental switching practically impossible. When toggled, the firmware announces LANG:en or LANG:bn over serial, and the Teacher Software updates its per-student language indicator automatically — with no configuration by the teacher required.
The physical Shift key (GPIO 32) acts as a modifier: holding it while pressing a Braille chord routes the lookup through a second set of dictionaries. A fixed priority order ensures predictable output: numbers first (Shift + A–J = 1–0), then math and punctuation symbols, then uppercase letters in English mode, or dependent vowel signs (কার) in Bangla mode. An entire layer of additional characters is unlocked without adding a single physical key to the device.
The firmware, ESP32, mechanical switches, 3D-printed enclosure, and Teacher Software are each independent components with clean interfaces between them. A broken switch is replaced by pressing in a new Outemu Red (~$0.38). A damaged ESP32 is swapped for a new module (~$5). A cracked enclosure shell is reprinted from the open-source STL file in a few hours. No single failure requires replacing the whole device, and no component is proprietary.
The Teacher Software is a single-page HTML/CSS/JS application that runs entirely in the browser using the Web Serial API. There is no backend, no database, no installer, and no configuration server. A teacher opens the file in Chrome or Edge, connects the USB cable, and the system works. Updates require only refreshing the page. The same file works for one student or an entire classroom — students are added as data rows, not as code changes.
Every character received from a student’s device is immediately appended to an in-memory data store with millisecond-precision timestamps. A 500 ms interval timer independently rebuilds all feed views directly from this stored history. Even if a DOM update is missed, the browser re-renders, or a visual glitch occurs, no character is ever lost from the record — because the store is the single source of truth, and the display is always rebuilt from it. The live view is a rendering layer on top of persistent data, not the data itself.
The novelty statement from the paper: “BrailleBridge does not replace Braille — it protects Braille as the student’s natural writing method while removing the barrier between Braille input and teacher-readable output.”