Driver drowsiness is one of the major problems that every country is facing. The ICT sector is continuously investing in the automaker industry worldwide to bring about digital transformation in existing vehicles and driving. The smart behavior of vehicles is becoming possible with the convergence of intelligent manufacturing, AI, and IoT. In this chapter, the authors are presenting a framework for efficient detection of driver's drowsiness by utilizing the power of deep learning technology. The use of convolution neural network (CNN) is explored, and the system is developed and tested using different activation functions. The proposed driver drowsiness framework is able to signify the drowsiness state of the driver and to automatically alert the driver. The accuracy of the proposed model is compared at different activation functions such as ReLu, SeLu, Sigmoidal, Tanh, and SoftPlus, and higher accuracy is achieved with ReLu as 98.21%.
TopIntroduction
Artificial Intelligence, Machine Learning, Deep learning, and IoT revolutionized the healthcare system and gave a new direction to the development of the nation. Disease classification, detection and prediction are increasing the expected life of human beings (Singh et al., 2021). On the other hand, people are losing their lives in road accidents for several reasons. There may be overspeeding, drunk and drive and drowsiness. Drowsiness means feeling unusually sleepy or tired during the day is typically recognized as drowsiness. Drowsiness is a cognitive function of the brain decreases when the person feels sleepy. The condition of drowsiness can be termed as a state of sleep deprivation and tiredness. This drowsiness can hinder the ability to perform any logical or complex tasks of a person. Drowsiness is really problematic and risky when the person is driving. Hence, it is very significant to detect drowsiness. As per our best knowledge, we do not have any concrete practices that convey a direct and clear result of the drowsiness condition.
According to the Central Road Research Institute (CRRI) study, 40% of highway accidents happen just because of drivers' drowsiness 1. According to a survey conducted by National Sleep Foundation, it has been brought in light that about 20% of drivers feel drowsy while driving vehicles (Dua et al., 2021). Real statistics could be much higher, however, as it is difficult to figure out whether a driver was drowsy at the time of the accident. A driver driving in a drowsy state is one of the key factors for road accidents. In “a car safety technology, driver drowsiness detection is essential to prevent road accidents. Nowadays, many people use automobiles for daily commutation, higher living standards, comfortability, and timing constraints to reach destinations. This trend leads to high volumes of traffic in urban areas and highways. In turn, it will raise the number of road accidents with several factors.” Primarily, Driver Drowsiness is the main reason for most road accidents. This tally can be lowered by alerting the driver of the drowsy state. Deep learning-based models have the potential to equip vehicles with various driver-assistance technologies.
The deep learning model outperforms in face detection (Hasan et al., 2021) with the ability to automate the feature detection that will be utilized in drowsiness detection through the closeness of eyes. Deep learning algorithms are categorized by the use of neural networks whose models are constructed of huge amounts of layers (Pouyanfar et al.,2018]. There is a specific type of deep neural network (DNNs) called convolutional neural networks (CNNs), which have great performance on computer vision due to its ability to detect the pattern and identify characteristics among images (Magan et al. 2022 and Luo et al., 2018). CNN is a kind of deep neural network mostly used to analyze visual imagery (Li et al., 2021). CNN is based on a particular type of method known as Convolution. Convolution is a mathematical process based on two functions that result from a third function which describes that the shape of one is changed by the other. The role of CNN is to minimize the image shape that can be easily processed by any digital device without losing the principal component required for getting a good prediction. Activation function determines what is to be fired or transferred to the next layer of neuron. Various activation function affects the performance of neural network model. Hence, it is important to identify which activation function is enhancing your system accuracy (Ertuğrul, 2018).