Department of Electrical and Computer Engineering

Department of Electrical and Computer Engineering

Blind Assistive Technology Bill Reading Device (BATBRD) Professor Aura Ganz

Ian McAlisterColin Smith

Chris NeylandErick Drummond

TEAM GANZ Preliminary Design Review

2Department of Electrical and Computer Engineering

OUTLINE

Background and Significance Ideal Final Product Requirements, Technical Challenges Design Alternatives Proposed Platform Imaging Device and Initial Testing System Specifications and Block Level Diagram Image Processing Algorithms Proposed MDR Prototype Specifications Conclusion and Pathway to FPR


Background and Significance

Population• 1.8 million legally blind individuals in the U.S. [1]

• 21.2 million reported experiencing vision loss [1]

Definitions• Legally Blind

• Visual acuity less than 20/200 in the better eye with the use of a correcting lens [2]

• Vision Loss• Reported unable to see even with corrections including

blind above

[1] http://www.lighthouse.org/research/statistics-on-vision-impairment/prevalence/

[2] 42 U.S.C. § 416(i)(1)(B) (Supp. IV 1986). http://www.law.cornell.edu/socsec/rulings/ssr/SSR90-05.html


Current Scenario• Interview with Jenny• Folding Schemes• KNFB Reader

Background and Significance (Cont)


Our vision: Bill Reading Wallet• Jenny’s Feedback and Reaction

Portable Low Power

• Day Long Availability On a Single Charge

Easy to Use Cost <$100 Bill Recognition Speed<2 sec

Proposed Ideal Final Product


OUTLINE



Product Requirements and Technical Challenges

Intuitive Interface design Form Factor Cost constraints Power considerations Performance Computer Vision algorithms HW/SW integration


Design Alternatives

Handheld Optical Scanner• Costly• Lack of Availability/Support

Hat Mounted Reader• Poor Form Factor• Numerous Image Processing Variables

Camera Based Contact Device• Small Form Factor• Reduces Image Processing and Lighting Variables• Easier to Integrate components


Platform ChoiceSystem Beagle Board Gumstix Earth + Summit daughter board FPGA

Approx Cost $150 $150 to $210 $30-$70CPU 600MHz 600MHz 60MHz or SlowerRAM 256MB 256MB low-power DDR RAM Up to 1.1 Mbit of On-

Chip Embedded Memory

Supported OS(s) Android, Angstrom Linux, and Win CE

Linux 2.6.29 No OS Needed

Size 3"x3" 3.15" x 1.54" x .16" Approx 1" x 1"Weight 1.41oz 3.57oz .357 to 1.41oz

Power 2W 1-2W Less than 1WDevelopment Community

Yes, Strongest Following Second Strongest Following Third Strongest Following

Libraries Binaries/Source Few Binaries, Source N/AReliability Flashable from home RTM for Re-flash Flashable From Home

DSP support Yes, Built In Yes, With Summit Board NoCamera & Other

PeripheralsMost Linux Compatible USB/RS232

Devices Will WorkLittle Documentation on Compatible

DevicesMinimal To No Documentation


Platform Decision: The Alternatives

FPGA• Small Development Community• Difficult to Program• Advised to Avoid by Field Experts

Gumstix• High Cost: >42% of budget• Specs Similar to Beagle Board• Less Documentation• Smaller Development Community


Platform Decision: The Beagle Board

Beagle Board• Low power consumption• Cost: <30% of budget• Form factor• Good processing speed• Strong development community• Simple integrated development environment


OUTLINE



Camera Requirements• Small Form Factor• Easy to Integrate

Device Alternatives• Discrete CMOS Camera

• Little Documentation• UART-Controlled Camera

• Slow• Problematic

Imaging Device


Imaging Device: PS3 EYE Webcam

Designed for Computer Vision Applications Hacking Tutorials Available Low-Cost High Speed


Initial Lighting Tests: Contrast

Red Ink

→ Best Contrast: GREEN

Conclusion: Use two colors (one image for each)

Green Ink

→ Best Contrast: RED


System Specifications: Bill Reading Device

The Bill Reader Device will have the Approximate Specifications:• 1 processing unit: Beagle Board

• Commercial Off-The-Shelf (COTS) Camera: PS3 EYE• Speaker• LED Lights• SD Card

• Low power usage• System standby when not in use


System Level Block Diagram


OUTLINE

Background and Significance Proposed Final Product Alternatives Requirements, Technical Challenges Design Alternatives Proposed Platform Imaging Device and Initial Testing Specs and System Block Level Diagram Image Processing Algorithms Proposed MDR Prototype Specifications Conclusion and Pathway to FPR


Image Processing Algorithms

Potential Algorithm (High-Level):

•Receive images (Red, Green)•Perform:

SmoothingThresholdingEdge Detection

•Compare statistically with database•(Rotate if necessary, recompare)•Vote on results


Evaluate image processing algorithms (9 NOV) Implement one image processing algorithm on a

PC (16 NOV) Learn hardware/software platform (20 NOV) Test the algorithm on ideal images (20 NOV) Build bill library (20 NOV)

All tasks critical foundation• Whole team is assigned to all tasks.

Proposed MDR Specifications


Mid February (CDR)• Complete and test image processing algorithm(s) on a

PC• Integrate and test hardware components• Finalize design

March• Integrate software to Beagle Board• Develop and fabricate project enclosure

April• Begin final testing and data collection• Analysis of data for presentation

Approximate pathway from MDR to FPR


Identified Current Issues Developed Idea with Ideal Design Determined an Attainable Project Design Decided on Specific Hardware for Our Project Highlighted Image Processing as #1 Challenge Developed/Assigned MDR Tasks around

anticipated Image Processing Difficulties

Conclusion


Questions ?


Colin: Intro, Schedule Ian: Research background and significance Erick: Camera selection and related research Chris: Platform selection research

Slide Credits: In Random Order

Documents

Department of Electrical and Computer Engineering