Article Preview
Top1. Introduction
Fatigue and drowsiness are one of the biggest causes for drivers and passengers, and that can lead to serious injuries, other accidents, and even death. Roundabout 3.28 million people die in highway crashes every year (NSC, 2020). The German Road Safety Council states that one-quarter of road deaths are due to the momentary fatigue of drivers. Researchers suggest that the drowsy driving fatalities are 350% greater than reported. Like in the year 2021, 58 % of drivers of the younger generation drove vehicles in sleepy and tired conditions, and 28% fell asleep (NSC, 2020). As stated by WHO (World Health Organisation), the cost of road traffic crashes in most countries is 3% of their gross domestic products (WHO, 2022). Mostly the major accidents from the insufficiency of alertness are caused by the unconscious mind or tiredness. As per the statement of NHTS (National Highway Traffic Safety), there are approximately one million accidents with 800 fatalities and 50K injuries per year in the US (NSC, 2020). According to a study report, it has been stated that the US Government spends 109 billion dollars a year and firms on accidents caused by drowsiness in drivers (NSC, 2020). Roundabout 93% percent of the world’s fatalities on the road occur in low or middle-income countries, even though there countries have 60% of worlds vehicles.
Behavioral interventions affect the visual acuity of the driver and are greatly influenced by lighting conditions, the efficiency of the calibration system, and some other external factors. If we are able to overcome lightning conditions, and by using this process, we can come up with a cheap and affordable solution for drivers that are based on the cloud. The variations in heart rate, brain waves, or electrical signals in body muscles come under physiological changes (Gupta, Agrawal, Yamaguchi, & Sheng, 2020). Although their interventions may come up with a precise indication of fatigue, things are greatly affected by the technical aspects. Prevention of huge numbers of accidents can be done if we will get successful in recognizing human activities but in modern days recognizing human action is a big challenge in the area of computer imaging. Achieving 100% accuracy is next to impossible but error needs to be minimized as much as possible. We haven’t reached a significant precision in detecting human behavior, especially in various environmental conditions. If a system/machine can detect drowsiness in human behavior, an alert can be sent to the driver and prevent accidents. There is also another way of detecting drowsiness based on behavioral data, physiological measurements, and vehicle-based data. Behavioral data includes the state of eye/ face/ head/ hand movements using a camera. Physiological measurements include EEG (Electro Encephalo Gram), heart rate, SRG, PPG, ECG (Electrocardiogram), and vehicle-based data captured from the pattern of the speed of the vehicle, braking system, deviation in lane position, and wheel motion (Awais, Badruddin, & Drieberg, 2017). But all these methods of detecting drowsiness have their own pros and cons. Getting physiological measurements is not fast enough to detect drowsiness in significant time, to alert the driver before any misfortune. For vehicle-based data, we need some specific costly equipment. We can consider the signs of drowsiness when there is continuously yawning, head movement, and constant blinking of an eye. Sometimes eye closeness for a long period of time can help to detect drowsiness. In recent years for detecting drowsiness, various facial features have been used such as inner brow and outer brow rise, lips movement for yawning, jaw-dropping, and head motions. Earlier, handcrafted methods were used to detect them. But it has a lot of drawbacks. One of them is when drivers are wearing sunglasses. In some practical situations, this method of detection becomes very ambiguous. To minimize this, people have begun to use machine learning (Deep Learning) methods to recognize these facial features.