Documente Academic
Documente Profesional
Documente Cultură
2010
AN ADAPTIVE DRIVER ALERT SYSTEM MAKING USE OF IMPLICIT SENSING AND NOTIFICATION TECHNIQUES
Gilbert Beyer, Gian Mario Bertolotti, Andrea Cristiani, Shadi Al Dehni
Agenda 1 2 3 4 5 Adaptation Strategy Sensor-Actuator Support Hardware and Software Performance Tests Discussion
Page 2 of 41
Driver Alertness
Page 3 of 41
Causes of Driver Distraction The driver himself (fatigue, drunkeness) The vehicle (type, condition) The environment (traffic, weather)
Page 4 of 41
Long-Term Driver Distraction Psycho-physiological state (fatigue) Secondary tasks (operating car stereo) Social activities (chatting with co-driver)
Page 5 of 41
Page 6 of 41
Sensing Driver Distraction Hands-off-wheel Time Eyes-off-road Time Conversational Level Emotional expressions Signs of drowsiness
Page 7 of 41
Notification Modalities 1 Vision, Hearing, Touch (each having its own timely behavior) 2 Multimodal information
Page 8 of 41
Adaptation Strategy Actuator: Sensor 1: Sensor 2: Sensor 3: Visual, implicit Drowsiness Head-position Can Bus data
Page 9 of 41
Combing Car and User Data Eyes-off-road while... 1 driving fast, is dangerous 2 standing, is safe 3 turning, is required
Page 10 of 41
Head-up Displays
http://www.bmw.com/ (2010)
Page 11 of 41
Dienelt, M. (2005)
Page 12 of 41
Page 13 of 41
Sensing Drowsiness 1 Intrusive techniques: Electromyogram, Respiration, GSR, Electro-oculography 2 Non-intrusive techniques: Vision-based
Page 14 of 41
Automotive Demonstrator
Page 16 of 41
Sensor-Actuator Support
Page 17 of 41
Simulator Setup
Page 18 of 41
Page 19 of 41
Drowsiness Detection
Page 20 of 41
Drowsiness Detection
Page 21 of 41
Main Constraints
Page 22 of 41
Page 23 of 41
Eye Detection Algorithm 1 2 3 4 Search eye using Viola-Jones If eye is found, save copy as template If eye is not found: template matching Successive failures: re-initialization
Page 24 of 41
Page 25 of 41
Blink Test Algorithm 1 2 3 4 Eye-image processed to binary picture Eliminating pupil reflections and noise Obtain vertical projection histogram Calculate eye openness level
Page 26 of 41
Greyscale image
Binary image
Page 27 of 41
Open eye
Closed eye
Page 28 of 41
Head-Position Detection
Page 29 of 41
Experiments and User Analysis 1 Performance Tests 2 Driver Observation in the Lab 3 Road Tests
Page 30 of 41
Passive Sensing Performance Tests Collected sets of data: 1 Recognition of slow/fast blinks 2 Tracking of slow/fast eye movements 3 Slow/fast head movements 4 Blink fail errors
Page 31 of 41
Passive Sensing Performance Tests Kinds of error: 1 tracking miss (in successive frames) 2 detection miss (shadows or occlusions) 3 blink miss (e.g. due to rapid blinks) 4 blink fail (erroneously detected blink)
Page 32 of 41
Page 33 of 41
Page 34 of 41
Page 35 of 41
Lab Study
Page 36 of 41
Lab Study 1 2 3 4 Eyes-off-road and Car State Symbols 15 Participants (students, colleagues) Co-Driver was investigator Semi-structured Interviews and Camera Observations
Page 37 of 41
Lab Study 1 80 % understood that displayed icons related to their driving behavior 2 66 % found notifications helpful to adapt their driving style
Page 38 of 41
Onboard Testing Session 1 Displaying Car State Information 2 Thinking Aloud Method 3 Refinement of the hardware and software setup
Page 39 of 41
Page 40 of 41
Discussion Algorithm Improvements Sensing Techniques Future Tests and Studies HUD Development System Design
Page 41 of 41