Wristband-Type Driver Vigilance Monitoring System Using Smartwatch IEEE SENSORS JOURNAL, VOL. 15, NO. 10, OCTOBER 2015 Boon-Giin Lee, Member, IEEE, Boon-Leng Lee, and Wan-Young Chung, Member, IEEE Presenter: Siao, Jing-Jhou Advisor: Dr. Pei-Jarn Chen Date:2015/12/16 1
Outline Introduction Methodology Experimental results Conclusion 2
Introduction 3
Driver vigilance Vehicle-based control behavior Steering wheel movement Physiological state Photoplethysmogram (PPG) Respiration signals 4
Steering wheel movement Smartwatch motion sensors Accelerometer Gyroscope sensors 5
PPG sensor A sport wristband with a Bluetooth low energy module Transmitted the PPG signals to smartwatch 6
Methodology 7
Wearable type monitoring system Sensor module (sensing and A/D convert) Smartwatch module (receive, analysis, information display, and alert) 8
Hardware specification 9
Wearable type monitoring system 10
Steering wheel angles (SWA) Pitch (rotate at x-axis) Roll (rotate at y-axis) Yaw (rotate at z-axis) 11
Wristband-type system Wristband-type system 12
Hands position regarding to the steering angle 13
14 Angle derivation
Photoplethysmogram 15
PPG-Derived Respiration 16
Features extracted from motion sensors, PPG, and PDR 17
Karolinska sleepiness scale (KSS) 18
Weighted fuzzy c-means (WFCM) Fuzzy c-means is an essential tool to find the proper cluster structure of given data sets in pattern and image classification A weighted fuzzy C-Means algorithm is proposed to improve the performance of both FCM models for high-dimensional multiclass pattern recognition problems 19
Weighted fuzzy c-means (WFCM) 20
Accuracy rate ( the age and gender) 21
Comparison of classifiers 22
Experimental Results 23
Driver Vigilance Monitoring System 24
Driver Vigilance Monitoring System 25
Conclusion 26
Conclusion This study presents the steering wheel movement derived from motion sensors is a great indicator to measure driver vigilance level. On the other hand,by integrating the driver physiological state into vigilance prediction, it can increases the true prediction rate (96.5%) as well as reduces the false prediction rate (4.2%). 27
Thanks for your attention ! 28