SENIOR DESIGN PROJECTS Design Review Eye Tracking System Optimization Zachary Harvey Piyush Agarwal Robert Laiacona Lowren Lawson
Presentation Overview Project Overview Zachary Harvey Typical Operation / Sensors Rob Laiacona System Overview Lowren Lawson TI Divinci DM355 / Possible Problems Piyush Agarwal Prior Work / Power Requirements Multidisciplinary Components / Societal Impact / Power requirements Testing / System Integration/Alternatives/Costs
EXISTING SYSTEM PC53XS scene camera PC206XP eye camera IR LED http://www.cis.rit.edu/pelz/publications/ETRA04_babcock_pelz.pdf
PROBLEMS WITH THE EXISTING SYSTEM Video Synchronization Video Cutout Data Processing New Eye Tracking Algorithm Power Management Monitoring Too Bulky Offline analysis restriction
CUSTOMER WANTS Basic Level More Advanced Features Synchronized 30 FPS video LCD screen display Four hour video storage Simple Interface 8 hours storage / battery More Advanced Features Control Wirelessly Storage via H.264 Video multiplexing
Typical Operation Subjects are taken outdoors / indoors in daylight Operation in fair weather Subjects are instructed to hike and climb
Sensors Cameras CMOS Analog NTSC encoded video stream ~30 FPS to capture all eye movement PC53XS scene camera PC206XP eye camera IR LED
Typical Data
System Startup Startup Self Test Enter standby mode Wait for input Determine number of cameras Determine battery power Storage Space LCD test Enter standby mode Wait for input
User Interface
HIGH LEVEL SYSTEM DESIGN
Leopard Board TI DM355 MPEG4 Coprocessor Arm GPP (270 MHz) TI DM365 MPEG4 / H.264
Problems Leopard Board interfacing Processing Expense Packaging
Problems (Leopard Board Interface) Digitizing two video streams Breakout board needed Getting drivers to work properly Power consumption Mitigation Switching platforms to a more powerful SBC
Problems (Processing Expense) MPEG4 / H.264 video compression is computationally expensive Hardware capable of processing is expensive SBC solution ~500 - 1000$ Mitigation Leopard Board (~100$) Limited to one stream / MPEG4
Problems (Packaging) System must be durable System must be low power Subjects are told to climb / hike outdoors Connections must be secure System must be low power System runs off of 7.2V 4000 mAH battery Heat dissipation Mitigation Locking connectors Low power parts Enclosure
PRIOR WORK Applied Science Laboratory EYE TRAC openEye Open Source Expensive openEye Open Source Poor recording capabilities RIT Multidisciplinary Project Poor software implementation Windows XP + Lab View More research needs to be done on what went wrong NASA eye tracking device Works specifically with their equipment Not commercially available
Power Requirements User would like 4-8 hours of operation Using 7.2 4000 mAH battery Leopard board will last 4-5 hours
Multidisciplinary Components Electrical Engineering Power consumption Imaging Science Computer Science Packaging Science
SOCIETAL IMPACT Visual Perception Research Commercial Advertising Medical Research Academic Strategies
Testing Procedure Video Stream Testing Power Consumption Testing Regression Testing Interface testing
System Integration Initially VGA camera module will be used for input Data will be recorded to SD card Once Video can be acquired, the NTSC streams will be used
Alternative Approaches SBC Easy but expensive Reverse engineer existing video capture device Fully Custom Platform
COST ESTIMATES Item Item Cost Expected Cost Item Qt. Eye-tracking Cameras, glasses, IR LED, and breakout box. (Headgear) $5000.00 $0.00 1 16 GB SD Card $35.25 2 Camelbak Backpack $62.50 Battery 7.4 V (4000 mAh) $50.00 Power Distribution Block $10.00 Microphone Cables LCD 6” Display $160.00 Video Decoder IC Audio Decoder IC Leopard Board $84.00 $5553.00 $289.25