Driver distraction detection and recognition using RGB-D sensor
URL: http://arxiv.org/abs/1502.00250v1
archive: archived pipeline: cataloged verified
Abstract
Driver inattention assessment has become a very active field in intelligent transportation systems. Based on active sensor Kinect and computer vision tools, we have built an efficient module for detecting driver distraction and recognizing the type of distraction. Based on color and depth map data from the Kinect, our system is composed of four sub-modules: eye behavior (detecting gaze and blinking), arm position (is the right arm up, down, right of forward), head orientation, and facial expressions. Each module produces relevant information for assessing driver inattention. They are merged together later using two different classification strategies: AdaBoost classifier and Hidden Markov Model. Evaluation is done using a driving simulator with 8 drivers of different gender, age and nationality for a total of more than 8 hours of recording. Qualitative and quantitative results show strong and accurate detection and recognition capacity (85% accuracy for the type of distraction and 90% for distraction detection).
Summary
Builds a Kinect RGB-D driver-distraction detection module combining four sub-modules (eye gaze and blinking, arm position, head orientation, and facial expression) whose outputs are fused with AdaBoost and Hidden Markov Model classifiers. Evaluated on 8 drivers and 8+ hours of simulator recording, the system reached approximately 90% accuracy for distraction detection and 85% for distraction-type recognition. The authors argue each sub-module is independently usable for related inferences such as fatigue.
Key finding
An RGB-D Kinect rig fusing gaze, arm-position, head-pose, and facial-expression cues via AdaBoost or HMM classifiers achieves ~90% distraction-detection and ~85% distraction-type-recognition accuracy in simulated driving.
Methodology
Computer-vision system development and evaluation. Color and depth maps from a Microsoft Kinect were processed through four feature sub-modules; AdaBoost and HMM classifiers were trained and tested on simulator recordings of 8 drivers (mixed gender, age, nationality), totalling more than 8 hours of data.
Sample size: N=8 drivers, >8 hours of simulator recording
Quality score: 5 / 5