EyePhone Activating Mobile Phones With Your Eyes Felix Nwaobasi - PowerPoint PPT Presentation

EyePhone Activating Mobile Phones With Your Eyes Felix Nwaobasi Felix Nwaobasi CS525: Mobile & Ubiquitous Computing

Mobile Interactions Mobile Interactions • Lots of different ways to measure Lots of different ways to measure user gestures on a phone (accelerometers cameras etc) (accelerometers, cameras, etc) • Touchscreen was a major • Touchscreen was a major advancement • Can we go further? 2 2 Worcester Polytechnic Institute

3 3 HCI vs HPI HCI vs HPI Worcester Polytechnic Institute

HCI HCI • Tends to focus on ideal settings Tends to focus on ideal settings • Takes full advantage of the a T k f ll d t f th computer’s computational power • External sensors used, often , encouraged 4 4 Worcester Polytechnic Institute

HPI HPI • Deals with users on the move under Deals with users on the move under different contexts & conditions • Huge concern regarding energy consumption and computational consumption and computational power • External sensors frowned upon 5 5 Worcester Polytechnic Institute

C Creating the EyePhone ti th E Ph 6 6 Worcester Polytechnic Institute

EyePhone Features EyePhone Features • Maps a user’s eye movement to a Maps a user s eye movement to a position on the display • Blinks from the user correspond to a click li k • Hands-free! 7 7 Worcester Polytechnic Institute

The EyePhone Algorithm The EyePhone Algorithm • Eye detection Eye detection • Eye template creation • Eye tracking E t ki • Blink detection 8 8 Worcester Polytechnic Institute

Eye Detection Eye Detection • Uses motion analysis operations on Uses motion analysis operations on consecutive frames • Focused on eye contours 9 9 Worcester Polytechnic Institute

Eye Detection (cont.) Eye Detection (cont.) 10 10 Worcester Polytechnic Institute

Eye Template Creation Eye Template Creation • Used whenever eye gets lost y g • Created first time user uses Created first time user uses EyePhone • Reduces computation time & preserves battery life p y • Not very effective if lighting changes 11 Not very effective if lighting changes 11 Worcester Polytechnic Institute

Eye Tracking Eye Tracking • Based on eye template matching Based on eye template matching • Correlation score between search • Correlation score between search window and open eye template • Use correlation coefficient to improve accuracy improve accuracy 12 12 Worcester Polytechnic Institute

Eye Tracking (cont.) Eye Tracking (cont.) • Correlation coefficient of 4 works Correlation coefficient of .4 works great for eye template matching 13 13 Worcester Polytechnic Institute

Blink Detection Blink Detection • Uses thresholding Uses thresholding • Created four different thresholds to • Created four different thresholds to account for phone’s bad quality camera camera • If correlation coefficients are within • If correlation coefficients are within these thresholds, then eye is closed 14 14 Worcester Polytechnic Institute

15 15 Evaluation ti l E Worcester Polytechnic Institute

Experiments Experiments • Daylight Exposure/Stationary Subject Daylight Exposure/Stationary Subject • Artificial Light/Stationary Subject 16 16 Worcester Polytechnic Institute

Experiments Experiments • Daylight Exposure/Person Walking Daylight Exposure/Person Walking • Distance/Tablet correlation 17 17 Worcester Polytechnic Institute

Experiment Results Experiment Results 18 18 Worcester Polytechnic Institute

EyePhone & HPI EyePhone & HPI • Lightweight application Lightweight application • Camera obtained 15 frames/sec • Application only runs when user is A li ti l h i looking at display • Three hours of battery life if used continuously 19 19 Worcester Polytechnic Institute

EyePhone & HPI (cont.) EyePhone & HPI (cont.) 20 20 Worcester Polytechnic Institute

Possible Applications Possible Applications EyeMenu EyeMenu • Maps eye position to 1 of 9 buttons • Blink to click! • Great for people with disabilities 21 21 Worcester Polytechnic Institute

Future Work Future Work • Increase battery life Increase battery life • Improve eye template creation • Improve eye template creation • Minimize false positives • Minimize false positives 22 22 Worcester Polytechnic Institute

23 23 ? Questions? ti Q Worcester Polytechnic Institute