PROJECT TITLE :

Driver Distraction Detection Using Semi-Supervised Machine Learning

ABSTRACT:

Real-time driver distraction detection is that the core to many distraction countermeasures and fundamental for constructing a driver-focused driver help system. Whereas information-driven ways demonstrate promising detection performance, a specific challenge is how to cut back the considerable value for collecting labeled data. This paper explored semi-supervised ways for driver distraction detection in real driving conditions to alleviate the value of labeling training data. Laplacian support vector machine and semi-supervised extreme learning machine were evaluated using eye and head movements to classify 2 driver states: attentive and cognitively distracted. With the additional unlabeled information, the semi-supervised learning ways improved the detection performance (G-mean) by zero.0245, on average, over all subjects, as compared with the traditional supervised methods. As unlabeled coaching data can be collected from drivers' naturalistic driving records with little further resource, semi-supervised methods, which utilize both labeled and unlabeled knowledge, will enhance the efficiency of model development in terms of your time and cost.