1. Introduction
Over the last few years, road accident incidents have seen a tremendous increase mainly associated with driver distraction caused by mobile phone calls. The U.S. federal government has reported [
1,
2] that more than 30,000 people are killed on U.S. roads every year in crashes related to distracted driving. Despite knowing the risk of using a mobile phone while driving, 70% of drivers use a mobile phone during driving. There have been recent studies showing the effect of popular social networking games, such as Pokémon GO, leading to many accidents, where users play games while driving [
3]. The chances of crashing become higher if mobile phones are used because this reduces driving performance. The control of the sidelong and longitudinal position of the car is one of the basic needs during safe driving; however, frequent use of mobile phones while driving results in poorer lane-keeping, slower response time, and more variable speed [
4,
5]. In addition, the distance between the driver car and the front car, the fact of wandering out of the driving lane, and a reduced awareness of surroundings are some factors that lead to frequent road accidents and mortalities. In order to control the abrupt use of mobiles phones while driving, we need a smart portable system, which can control and monitor the behavior of a driver whenever the distracted driving exceeds a threshold and the risk of road crashing becomes imminent.
The consequences of using a mobile phone during driving are more lethal than communicating with fellow passenger, and research has shown that drivers busy in discussions over the phone have missed highway exits four times more frequently than those talking with passengers. The drivers talking with travelers did not demonstrate any significant differences with the lone drivers in the simulator environment [
6]. The use of cell phones for chatting, messaging, playing media, web browsing, gaming, using the global positioning system (GPS), or working other telephone applications or apps is a dangerous act leading to distracted driving and road crashes [
6]. This was evident from a report in 2010 from the U.S. National Highway Traffic Safety Administration (NHTSA) stating that 995 drivers died only because of distraction caused by the use of phones. Similarly, another study in March 2011, carried out by a U.S. insurance agency, State Farm Insurance, declared that 19% of the drivers involved in road accidents were busy talking on a cell phone while driving [
7].
The tendency of mobile phone use during driving is more frequent in youth, with one of the studies reporting that more than 90% of college students were involved in initiating, reading, or replying to messages while driving. Messaging while driving, is considered the most harmful among all types of distraction using a mobile phone while driving, and it has been reported that there is a six-fold increase in distraction-related crashes due to text messaging [
7]. Mobile phone texting, using MP3s, and other distractions may hinder the capacity of young drivers to control the vehicle and their ability to anticipate and manage hazards [
8]. However, collision avoidance systems, electronic stability control, vehicle tracking systems, and intelligent speed adaption may help to reduce the problem even though technology alone cannot make the young driver safe.
A study [
9] was conducted that was based on a self-reported questionnaire survey carried out among 242 young drivers in Riyadh, Saudi Arabia to obtain detailed insights into traffic violations committed by young Saudi drivers. The study showed that excessive speeding, which is mainly caused due to running late or testing a car’s performance, is the leading cause of traffic accidents and traffic violations. Moreover, driving very close to the front car, which inhibits the driver to stop in an emergency, is another significant factor leading to traffic accidents. A study was carried out at the American University of Beirut (AUB), Lebanon and at George Washington University (GWU), United States of America to investigate the differences in red-light violations and driving behavior of drivers in those two countries. It has been reported that AUB students engage more in dangerous driving behavior than GWU students do, whereas GWU students are prone to violate traffic rules and red-light signals in the simulator [
10]. A study was carried out on 83 new license holder young drivers in private cars, where the system was acquiring driving performance including secondary task engagement and driving environment logging. This study showed that teenage drivers are frequently engaged in secondary tasks and tend to regulate themselves poorly based on the surrounding environmental conditions. Moreover, the teenagers are greatly influenced by peers with respect to engaging in secondary tasks [
11]. A system was demonstrated in the study [
12], which merged the driver’s background data and driving data to assess the good/bad driving pattern. The score for driving performance managed through the in-vehicle smart system, which provided feedback to the drivers to improve their driving, was found to be useful for taxi companies. There have also been studies where the authors have tried to detect driver distraction using semi-supervised machine learning without developing a prototype [
13].
Various technologies and smartphone applications have been developed in recent years in order to limit the frequent use of mobile phones during driving. One of these is Google Glass, but it is still not safe to use [
14]. In this same vein, a three-axis accelerometer of an Android-based smartphone was built up with multiple sensors to improve a driver’s awareness to maximize safety [
15]. A hardware device that can detect mobile phone use while driving and later block mobile communications could be a very fruitful option for monitoring and controlling road accidents. For example, radio frequency identification (RFID) technology could be used to record the data and send the vehicle’s plate number to a control center when the driver uses the mobile phone and a radio frequency (RF) blocker can be used to block the mobile phone. However, the regulatory commission of some countries does not allow an RF blocker or jammer to be implemented in the car. The use of smartphone accelerometer sensors is another important technique to monitor vehicle status that involves the application of a principal component analysis (PCA) algorithm with time, frequency as well as power spectral density features of the sensor data to predict the vehicle status. This mobile sensor proved beneficial in identifying driving behavior using mobile phone applications [
16]. However, this requires a high-performance computational capability in the smartphone application, or the application must be implemented in the cloud. The low-speed following mode (LSM) uses millimeter wave radar to identify the speeding up, deceleration, and stopping of the front car to appraise the distance from the front car; in the interim, the driver likewise controls the brake and the fuel systems to keep up the vehicle distance within the safety range. When the front car encounters a strange condition, the system produces an alert sound to warn the driver [
17]. However, this assistive technology is implemented in some expensive cars and is not easy to implement in all available vehicles. The Lane Keeping Assist is a useful system to monitor the passing or separating lane by using a camera fixed to the front of the vehicle. The system produces a cautioning signal when the driver crosses or enters the opposite side of the passing line without using the correct light direction indicator [
17]. However, this camera-assisted system is prone to making mistakes during rough weather conditions and bad road conditions, and the image processing requires a powerful computer to be installed on board. In a very recent work [
18], a content analysis was conducted on 29 English smartphone applications to identify the stopping, preventing, or reducing phone use behavior while driving, detected by the applications. The functionality of these applications was determined based on application–mobile phone interaction, application–driver interaction, and application–context interaction. Most of these applications focused on blocking specific phone functions; however, the applications did not focus on simplifying phone tasks while driving and none of them was designed to study driving behavior. Another recent work by Delgado et al. [
19] showed that the strongest perceived benefit of cellphone blocking apps was decreasing distraction (86%). The predominant reason among young drivers for not wanting to use this technology was not wanting parents to monitor their behavior (60%). This work shows the importance of developing driver-friendly applications while controlling a driver’s mobile usage. The systems that have been reported so far in the literature did not present a feasible solution that could acquire the driving behavior from the car and use it to control the mobile phone automatically.
In this work, we have proposed a hybrid (hardware and software combined) solution to monitor driving behavior and keep track of a driver’s mobile phone usage, and to control the mobile phone call when the car speed reaches a certain threshold. The on-board diagnostics (OBD)-II port of the vehicle was used to get the vehicle’s real-time data. It was used to obtain the car speed very accurately, the accelerometer (ACM) sensors were used to identify some aspects of the driving behavior, and a mobile application written in the Android platform was used to monitor, log, and report driving behavior and control mobile phone calls. We decided on several driving behaviors to be studied and monitored in this work, based on an anonymous self-reported survey that was conducted with 600 male and female teenager drivers belonging to different nationalities. The survey was done in order to determine the pattern and frequency of mobile phone use while driving, and to find out about driving behavior, distracted driving due to mobile phones, and drivers’ level of recommendations regarding the use of technologies to assist drivers or control mobile phones in Qatar.
3. Analysis
In this work, survey data analysis was accomplished in Microsoft Excel 2016 and vehicle data were initially analyzed in MATLAB 2018 and later (after the development of the mobile application) done in the mobile phone. Initial development and testing of the smartphone application were carried out on a Samsung Note 8 mobile phone, which is powered by an Exynos 8895 Octa-core processor, along with 6 GB of RAM and 64 GB internal memory. The operating system installed on the phone was Android 7.1.1 (Nougat) and enabled with Bluetooth Low Energy (BLE) 5.0. However, the smartphone application was tested in several lower-end smartphones in the testing phase.
Survey Analysis: Detailed chi square statistics [
21] were performed on two major questions, namely, (i) Is it also a good idea to restrict texting, gaming, or browsing while driving and (ii) Is it a good idea to restrict phone calls to only emergency family contacts while driving?, to check the accuracy of the predictions made by the authors based on the literature review. The first question had the assumption that all participants would state that it was a very good idea. For the second question, the authors assumed that all participants would state that it was not safe.
Preliminary Analysis in MATLAB: Initial accelerometer results were smoothened by the moving average filter in MATLAB and averaged over trials and subjects to calculate the mean of the x-, y-, and
z-axis data (
Figure 10). It was observed that, for this work,
z-axis data were not useful and therefore not used for further processing. Moreover, engine rpm had an offset value when the vehicle was started and changed from that reference and the variation reflected in the engine rpm was also reflected in the speed and therefore rpm was not used for calculating behavior. The accelerometer and OBD-II module data were analyzed to see the trend of driver behavior in relation to the nature of the x and
y-axis and speed data.
Data Analysis in Mobile Phone: The mobile application used here to collect the vehicle data from the OBD-II device and to store them temporarily in the mobile memory until they were processed to make a decision (i.e., the x, y, and speed data) were buffered in the mobile memory. The buffering duration was kept to 10 s to get enough data to observe the changes while not cluttering the memory of the mobile phone. The algorithm of real-time peak detection is very robust because it constructs a separate moving mean and deviation for the buffered data, such that signals do not corrupt the threshold. Future signals are therefore identified with approximately the same accuracy, regardless of the number of previous signals. The algorithm takes three inputs: lag represents the lag of the moving window, threshold represents the z-score at which the algorithm generates peak, and influence represents the influence (between 0 and 1) of new signals on the mean and standard deviation. For example, a lag of 5 (moving window) will use the last five observations to smooth the data. A threshold of 3.5 (estimated from MATLAB study) will signal if a datapoint is 3.5 standard deviations away from the moving mean. In addition, an influence of 0.5 gives signals half of the influence that normal datapoints have. Likewise, an influence of 0 ignores signals completely for recalculating the new threshold. An influence of 0 is therefore the most robust option; putting the influence option at 1 is least robust.
In the mobile application, there were three subclasses: two for x and y data analysis and the third for the speed data analysis. The subclasses for the x and y data analysis helped to identify the peak of the x and y movements of the vehicle which, in turn, helped to identify normal and abnormal behaviors of the driver, whereas the other subclass sent responses based only on speed data. The filtered speed data were sent to another class, which made a decision based on speed, that is, the vehicle was either in driving mode or stopped, and sent a callback to this class on the current status. Based on the status of the vehicle, the control function started tracking or stopped tracking and the calling function was also deactivated or activated.
Figure 11 shows the detailed stages of how the mobile application was designed to log normal or abnormal behavior.