The Hybrid Feature Selection Method to Recognize Driver’s Inattention Issue

Abstract

Driver inattention detection has gained more attention in recent decades and the literature shows evidences of various Driving Monitoring and Assistance Systems. Those systems are aimed to assist the drivers to improve their performance and to avoid road accidents. The driving monitoring systems keep on the driving position of a driver and to deliver required support for comfortable and safe driving. A classification technique along with more number of features would be a comprehensive standard in this backdrop. Nevertheless, if the count of features are extremely high subsequently the intricacy of training phase would upsurge over and above there is a venture that the classification method might be contrary. This research work proposes a hybrid metric to estimate more robust feature score for preliminary feature selection. Moreover, the metric could be used to reduce the feature space both in horizontal as well as in vertical direction. The same metric has been used to identify the relevant features as well as to remove the redundant samples. The performance of the feature selection step is evaluated with Support Vector Machine (SVM) classifier.