Article Preview
TopIntroduction
This paper applies a few machine learning mechanisms to detect whether or not a driver is drowsy. A drowsy driver will likely increase the chances of accidents that result in injuries with possible fatalities. A driver is likely to fall asleep at the wheels while driving, possibly after too short a sleep, altered physical condition, or during long journeys. While in a drowsy state, a driver's level of vigilance diminishes, thereby increasing the probability of occurrence of accidents. Driver drowsiness is an important factor that contributes to the total number of deaths and fatal injuries due to traffic accidents globally. It is estimated that 30% of all traffic accidents have been caused by drowsiness. It was demonstrated that driving performance deteriorates with increased drowsiness with resulting crashes constituting more than 20% of all vehicle accidents (Fuletra, J. D., 2013). National Highway and Traffic Safety Administration (NHSTA) reported that, in the year 2010, traffic accidents 2.4 million vehicles in the US cost $871billion and resulted in 33,000 deaths and 3.9 million injured (US DOT NHSTA 2015).
The National Sleep Foundation reports that 60% of adult drivers had driven a vehicle while feeling drowsy and 30% had actually fallen asleep (National Sleep Foundation). Hence, it is important to install in cars a real-time drowsiness monitoring system that can help in preventing traffic accidents from occurring. This paper develops a system that can applied to controlling accidents due to drowsiness. This paper presents a literature review of driver drowsiness detection based on behavioral measures using machine learning techniques. Facial features contain information that can be used to interpret levels of drowsiness. Many facial features including eye blinks, head movements and yawning provide crucial information concerning the level of drowsiness of a person. This paper compares the results produced by various machine learning techniques such as SVM [Support vector machine], HMM [Hidden Markov Model] and CNN [Conventional neural Network]. The results produced by CNN are superior to the results produced by the other two (Ngxande, M. et al 2017).
(Srijayathi, K. & Vedachary, M. 2013) implements a module for Advanced Driver Assistance System (ADAS) to reduce the number of accidents due to driver’s fatigue and hence increase the transportation safety. Here in (Karmakar, S. et al 2018) the system deals with automatic driver drowsiness detection based on visual information and Artificial Intelligence. It proposes an algorithm to locate, track, and analyze both the drivers face and eyes to measure PERCLOS, a scientifically supported measure of drowsiness associated with slow eye closure. (Eddie, E. et al 2018) reports on the implementation of a closed-circuit television to the driver supported by artificial vision techniques and enforced with a smart phone so as to observe and alert once somnolence signs are detected. In this system, the authors have developed an algorithm for somnolence detection and an interface to display on the smart phone the state of somnolence. The system, in conjunction with a smart phone, acquires images, detects facial features, estimates physiological parameters, determines the level of drowsiness, and alerts the driver with aural and visual alarms.
The authors in (Sharayu, R. et al 2015) describe a system that comprises of a couple phases to detect whether a driver is drowsy. Specifically, the system tracks the driver’s facial expressions, detects the contour of the eyes and the mouth areas and counts the rate of change of the contours. When the system detects fatigue in the eyes or the contour of the mouth points to yawning, the system alerts the driver. A video camera, placed underneath the front mirror, continuously records the driver’s face to detect whether the driver yawns. The primary step here is to monitor and track the facial expressions by analyzing the series of frame shots taken by the camera. Then the movement of the eyes and the mouth are analyzed. The analysis distinguishes between eye closure due to drowsiness and that due to yawning. However, this paper does not apply any machine learning techniques which are capable of detecting the facial features with improved accuracy. In (Assari, M. A. & Rahmati, M. 2011), the authors describe the implementation of a smart phone-based system similar to that in (Eddie E. et al 2018).To develop a non-intrusive system which may discover fatigue of the driving force and issue a timely warning, the system described in (Rusmin, P. H. et al 2013) employs a technique that monitors the driver’s eyes and detects symptoms of driver fatigue early enough to warn the driver and avoid accidents.