A Portable Device forOptical Recognition of Braille

Iain Murray

Curtin University of Technology

Introduction

• Purpose of the device

• A brief introduction to the Braille system

• Scanner operation

• The Prototype

• Image capture & processing

• Proposed improvements

• Conclusion

Why?

• Primarily for non Braille users– Education– Workplace

• Reproduction of worn or old Braille

• Electronic storage of Braille texts

The Braille System

• Tactile medium consisting of 6 “dots”

• Context sensitive– people = p– wh = : or which or : (colon)– Includes many rules for use

• Similar to shorthand

• May be printed both sides - interpoint

Scanner Operation

• Hand held portable unit

• Camera assembly is moved over the Braille line

• Operates in real time

• Results of the scan are displayed on the host computer

The PrototypeBlock Diagram

Image Capture

The Lens

CCD Operation

• 10ms integration time

• Pixels clocked out in 2.6ms

• 100 µs per sample

• Allows 7ms per slice processing time– (context switching and overheads considered)

CCD Operation

• SI pulses at 10ms intervals

• Pixels clocked out at 50KHz

Linear Motion Detection

• Slotted disk and IR detector/emitter

• Gear system to produce pulses at 200DPI

• Each pulse sends an interrupt to the DSP

Linear Motion Detection

Codewheel Gears

Illumination Array

Lens Housing

Selfoc Lens Array

Prototype Development Platform

• TMS320C50 DSP by Texas Instruments– Low cost– 10K words of data/program memory– Incorporates Analog Interface Circuit (AIC)

• TLC32040

• Programmable

• Include anti-aliasing bandpass filter

– Emulates serial (RS232) communications

Captured Image

Cell Element DetectionAre dots present?

• 25 samples at 100u second intervals

• Reference levels are averaged

• Samples are compared to reference level

• Results are written to scan word

Cell Compilation

• Scan words are captured at 200 dots/inch

• Stored as a single word, position oriented– 10 = bright– 01 = shadow– 00 = reference level– 11 = status

• dots are 1mm diameter => 8 scans per dot

Cell Compilation

• 50 samples kept (scan word) in the image buffer.

• Cells assembled from this image– Image buffer is compared to each possible

“ideal cell” (63 combinations)

Cell Compilation

• When a cell is matched:– Send binary Braille code

• Determined from a lookup table

– Clear the image buffer– Host processor “expands” grade 2 code

• Text is now available for display or storage

Grade 2 Decomposition• Algorithm by Paul Blenkhorn

• A simple state machine

• Allows for exceptions

• Uses look up tables and therefore allows :– Foreign language support– Special purpose codes e.g.

• Music

• Math

• Implemented in C

Proposed Improvements

• Quadrature motion detector will allow:– Reversal of scan direction– Decreased resolution

• Implement Fuzzy “dot” detection

• Use the SI pulse for triggering illumination

• Improve ergonomics

Conclusion

• Achieved– Image capture of a tactile medium– Dot recognition (position)– Grade 2 Braille decode– OCR of embossed Braille

Conclusion (continued)

• Further development should yield a useful device that overcomes the communication barrier between Braille users and the mainstream community.

The Prototype(and the development environment)