Cognitive Fatigue Detection in Vehicular Drivers Using K-Means Algorithm

In vehicular drivers, the cognitive fatigue has been one of the major factors that leads to loss of lives or disabilities due to vehicular accidents (Bundele and Banerjee in Proceedings of the 11th international conference on information integration and web-based applications and services, ACM, pp 739–744, 2009 [1]). The factors governing driver fatigue such as monotonous driving, traffic conditions on road, road conditions, insufficient sleep, anxiety, health conditions, work environment, type of vehicle, and driving comfort do affect largely the driving behavior. Many researchers across the world are working finding best suitable methods to minimize vehicular accidents. In this paper, our proposed approach is an alternative solution to detect cognitive fatigue in vehicular drivers to decrease the number of incidence of vehicular accidents specially those occurred due to cognitive fatigue (Hu and Zheng in Expert Syst Appl 36:7651–7658, 2009 [2]). The objective of this work has been to provide simple classification technique using K-means algorithms. The basic K-means and two modified versions have been proposed and validated to reliably detect cognitive fatigue while driving. It is of paramount importance that the sensory parameters are chosen such that they could be used without causing discomfort to the driver and without creating obstruction while driving. The data for simple physiological signals such as skin conductance (SC), oximetry pulse (OP), and respiration (RSP) for pre- and post-driving state of drivers has been used for sensing change in fatigue level of the drivers (Bundele and Banerjee in 2009 second international conference on emerging trends in engineering & technology. IEEE, pp 934–939, 2009 [3]). All features of statistical and wavelet were extracted and analyzed (Brown in D1 Methodological issues in driver fatigue research. Fatigue and driving: driver impairment, driver fatigue, and driving simulation, p 155, 1995 [4]). Selected features were used as input to the classifiers designed and implemented using basic K-means and two modified versions. Finally, comparative performance analysis of classifiers and the features is discussed. This paper discusses prominent results obtained during experimentation. Further the maximum classification accuracy could be produced by these features, so as to reduce the computational complexity (Bundele and Banerjee in Proceedings of the 11th international conference on information integration and web-based applications and services, ACM, pp 739–744, 2009 [1]). It could be found that a smaller set of features could provide the correctness of fatigue detection.