MDIT POLYTECHNIC COLLEGE

ULLIYERI,KOZHIKODE-673323

Seminar On
Enhancing Road Safety: Detecting Texting Distracted
Driving with Eye-Tracking and Machine Learning
DATE:12/07/2024
GUIDED BY:
Ms.Ashitha KS SUBMITTED BY:
LECTURER Abdul Basith VK
DEPT OF CT, M-DIT PTC REG NO: 2201131403
 CONTENTS

• Abstract
• Introduction
• Methodology
• Advantages
• Disadvantages
• Future scope
• Conclusion
• Reference
 ABSTRACT
Driving while texting is risky as it diverts the driver’s attention from the road and requires
them to shift between Handling their phone and the vehicle. Despite its dangers, many still
engage in this behaviour. To address this, some Companies have implemented features to
detect when someone is driving and restrict calls and messages unless confirmed Otherwise
by the user. This research introduces a method that combines eye tracking technology with
machine learning to Identify when a driver is texting. A driving simulator was utilized to
evaluate 26 participants under various conditions: normal Driving, emotional distraction,
cognitive distraction, and texting. Features were extracted from the eye movement data,
Encompassing fixation count and duration. After processing this data through machine
learning models, an impressive A driving simulator was utilized to evaluate 26 participants
under various conditions: normal Driving, emotional distraction, cognitive distraction, and
texting
.Accuracy rate of over 90% for identifying texting while driving. These findings are
promising and hint at the potential for a Real-time system that can detect and warn drivers
when they are engaged in texting.
 INTRODUCTION
Texting while driving is one of the common distractions that can potentially result in fatal
accidents and near-death experiences. Many studies have documented that distracted
driving, including texting, significantly increases the chances of accidents [1][3][7][8]. This
finding has prompted the prohibition of cell phone usage while driving in many states
across the United States. A recent report from the National Highway Traffic Safety
Administration (NHTSA) reported that using a cell phone while driving creates enormous
potential for deaths and injuries on the roads in the United States [7]; in 2020 alone, a total
of 3,142 people were killed in motor crashes involving distracted drivers. One potential
solution to mitigate the risk of distracted driving is to identify such behaviour as early as
possible in real time. Different machine learning algorithms have been used to predict driver
distraction.
A few are mentioned here: Kircher & Ahlstrom used logistic regression to predict driver’s
distraction based on the driving performance indicators [4]. Ragab et al. used random forest
to detect the driving distraction [9]. Liang et al. proposed a Realtime system for driver’s
cognitive distraction using support vector machines [5]. The same group proposed a
Bayesian networks-based distraction detection system [6]. These approaches mainly
focused on detecting visual distractions, such as adjusting in-vehicle devices (e.g., tuning
radio) or detecting cognitive distractions like cell phone conversations while driving.
However, less attention has been given to the automatic detection of texting distraction,
primarily a combination of visual distraction (eyes off the road) and sensorimotor
distraction (one hand moving between the car controls and the smartphone).
This paper addresses the research gap in detecting texting distractions automatically through
eye-tracking data. Most research work that reviewed the eye-tracking approach focused on
various driving distractions but not specifically texting distractions [11][12][13][14].
Recrate and Nunes reported the impact of verbal and spatial imagery tasks, such as
repeating words and mental image rotation, on eye fixations while driving [11]. Strayer et
al. used eye-tracking data to study the effect of hands-free cell phone conversation on visual
attention [12]. Sodhi and Reimer analyzed fixation durations for radio-tuning and rear-view
mirror-checking tasks [13]. Victor et al. studied eye movement data for visual and auditory
in-vehicle tasks [14]. Bit kina at el. suggested that the driver’s fatigue and workload is
based on gaze behavior and eye tracking metric could be utilized for predictions [16] Thus,
detecting texting distracted driving in real time remains an open challenge.
This paper aims to contribute to this area of research. A machine learning-based approach is
proposed in this work, which first extracts multiple features from the drivers’ eye-tracking
data, examines the features for statistical significance, and finally feeds the features that
pass the significance test to a set of machine learning algorithms. The following sections
explain the approach taken to answer this.
 METHODOLOGY
1. Literature Review
• Conduct a thorough review of existing research on distracted driving, particularly focusing on studies that use
eye-tracking and machine learning.
• Identify the gaps in the current research and define the specific objectives of your study.

2. Data Collection
• Participants: Recruit a diverse group of participants to ensure a wide range of driving behaviors.
• Equipment: Use high-quality eye-tracking devices (e.g., Tobii, SMI) to capture precise eye movements.
• Driving Simulation: Create a realistic driving simulation environment. This could be a controlled laboratory
setup or an on-road study with appropriate safety measures.
• Texting Task: Design a standardized texting task for participants to perform while driving. Ensure the task is
representative of real-world texting behaviors.
3. Eye-Tracking Data Acquisition
• Record eye movements, gaze points, fixation durations, saccades, and pupil dilation while participants drive and text.
• Simultaneously collect contextual data such as vehicle speed, lane position, and steering wheel movements.

4. Data Preprocessing
• Cleaning: Remove any noisy or irrelevant data from the eye-tracking and contextual datasets.
• Segmentation: Segment the data into smaller time windows (e.g., 1-second intervals) for analysis.
• Feature Extraction: Extract relevant features from the eye-tracking data such as:
• Gaze fixation points
• Fixation duration
• Saccade amplitude and duration
• Blink rate and duration
• Pupil dilation
• Labeling: Label each segment as "texting" or "non-texting" based on the task performed during that time window.
5. Machine Learning Model Development
• Data Splitting: Split the dataset into training, validation, and test sets.
• Model Selection: Choose appropriate machine learning algorithms (e.g., Random Forest, SVM, Neural
Networks) for classification.
• Training: Train the models using the training set. Tune hyperparameters to optimize performance.
• Feature Importance: Analyze which features are most indicative of texting while driving.

6. Model Evaluation
• Validation: Use the validation set to fine-tune model parameters and avoid overfitting.
• Testing: Evaluate the model on the test set using metrics such as accuracy, precision, recall, F1-score, and
ROC-AUC.
• Cross-Validation: Perform cross-validation to ensure the robustness of the model.
7. Real-Time Implementation
• Develop a real-time system for detecting texting while driving using the trained model.
• Integrate the system with vehicle infotainment systems or driver assistance systems to
provide real-time alerts.

8. Ethical Considerations and Privacy

• Ensure that the study complies with ethical standards and participants' privacy rights.
• Obtain informed consent from all participants.
9. Reporting and Dissemination
• Document the methodology, results, and conclusions in a detailed research paper.
• Present findings at conferences and in relevant journals to contribute to the field of road
safety.

10. Future Work

• Identify limitations of the current study and propose future research directions.
• Explore the potential for integrating additional sensors (e.g., EEG, heart rate monitors) to
enhance detection accuracy.
 ADVANATGES
1. High Accuracy
• Precise detection of distracted driving through detailed analysis of eye movements.
• Sophisticated machine learning models capture complex patterns.

2. Real-Time Monitoring
• Immediate feedback and alerts to the driver.
• Continuous monitoring of the driver’s attention.

3. Non-Invasive
• Minimal disruption to the driver’s natural behavior.
• Comfortable and unobtrusive eye-tracking devices.
4. Comprehensive Data Collection
• Rich datasets including gaze direction, fixation duration, and saccades.
• Integration with other contextual data for improved accuracy.

5. Personalization and Adaptation

• Models can adapt to individual driver behaviors.
• Continuous learning and improvement over time.

6. Enhanced Road Safety

• Helps prevent accidents caused by distracted driving.
• Supports law enforcement and policy-making with objective data.
7. Scalability
• Applicable to various vehicle types.
• Integration with other driver assistance technologies.

8. Cost-Effectiveness
• Reduces costs associated with accidents.
• Improves monitoring efficiency.

9. Advancing Research and Development

• Advances understanding of distracted driving.
• Encourages innovation in road safety technologies.
 DISADVANTAGES
1. Privacy Concerns
• Monitoring a driver’s gaze can invade their privacy and may lead to discomfort and
resistance from users.

2. High Cost
• Eye-tracking technology and machine learning systems are expensive to develop and
implement, potentially limiting widespread adoption.

3. Complexity
• Integrating eye-tracking systems with existing vehicle technology poses significant
technical challenges.
4. Accuracy Issues:
• Variations in eye movements, lighting conditions, and the use of eyewear can reduce the
reliability of the system in detecting texting behavior.

5. Driver Distraction
• The presence of eye-tracking devices and alerts from the system might themselves distract
the driver.

6. Ethical and Legal Issues

• There are concerns about surveillance, obtaining proper consent, and the potential misuse
of the data collected by the system.
7. Resistance to Adoption
• Drivers and car manufacturers may resist adopting such technology due to privacy
concerns, intrusiveness, and additional costs.

8. Data Security
• Ensuring the security of the collected data is crucial to prevent breaches that could expose
sensitive information about driver behavior.
 FUTURE SCOPE
The future scope for detecting texting while driving using eye-tracking
and machine learning is promising and multifaceted. As technology
advances, more sophisticated machine learning algorithms will likely be
developed, improving the accuracy and reliability of detecting distracted
driving. Integrating eye-tracking technology with semi-autonomous and
autonomous driving systems could significantly enhance overall vehicle
safety by providing real-time monitoring and intervention capabilities.
To address privacy concerns and gain user trust, stronger privacy
protections and data security measures will be essential. Technological
advancements are also expected to reduce the cost of eye-tracking
hardware, making it more accessible for widespread use. Regulatory
support from governments and safety organizations could further drive
the adoption of these technologies by mandating their use in new
vehicles.
Design improvements will focus on creating less intrusive, more driver-
friendly eye-tracking systems that minimize distractions and are easier
for drivers to use. Additionally, personalized feedback systems could
provide real-time alerts and assistance to help drivers correct distracted
behaviors, enhancing the overall driving experience.
The data collected by these systems can offer valuable insights into
driver behavior, which can be leveraged for in-depth analysis and to
inform future safety innovations and policies. Beyond the automotive
industry, there is potential for cross-industry applications of this
technology, such as in aviation, rail, and heavy machinery operation,
where maintaining operator attention is critical.
Furthermore, the insights gained from eye-tracking and machine
learning systems can be used to run educational campaigns about the
dangers of distracted driving, helping to raise awareness and promote
safer driving habits. By exploring these avenues, the effectiveness and
adoption of eye-tracking and machine learning technologies for
improving road safety can be significantly enhanced.
 CONCLUSION
This paper presents an eye movement analysis for detecting texting and
distracted driving. Three types of eye movement parameters (fixation
count, fixation duration, and saccadic duration) were analysed. The
parameters were compared against everyday driving (without texting);
the participants significantly altered their fixation patterns while texting
(p < 0.01). The cognitive and emotional distractionsdo not exhibit such
a dramatic change in the fixation patterns. The texting distraction affects
driving performance, too. The study results reveal that participants
significantly altered their driving by lowering driving speed and
acceleration to compensate for more considerable lane departure, thus
decreasing the risk of an accident.
Furthermore, machine learning algorithms were explored to classify
texting distracted drives with 90% and above accuracy. This allows the
development of a feedback system to alert drivers in realtime about the
possible danger of texting and take away some driving controls if
texting continues despite the alert. Of course, more research is
necessary to handle this alteration of machine-human interactions in
realtime. Since eye fixation data offers discriminatory solid power, the
next step is to extend this offline analysis to a realtime analysis.
The development of a machine learning classifier that can predict
texting distractions as early as possible could be facilitated by using a
mere few seconds of eye gaze data from the start of texting
 REFERENCE
[1] Simone Benedetto et al., “Driver Workload and Eye Blink Duration,” Transportation Research Part F: Traffic Psychology and
Behaviour, vol. 14, no. 3, pp. 199-208, 2011. [CrossRef] [Google Scholar] [Publisher Link]
[2] Tanja Blascheck et al., “State-of-Theart of Visualization for Eye Tracking Data,” Eurographics Conference on Visualization, pp. 1-
20,
2014. [CrossRef] [Google Scholar] [Publisher Link]
[3] Jaeik Jo et al., “Vision-Based Method for Detecting Driver Drowsiness and Distraction in Driver Monitoring System,” Optical
Engineering, vol. 50, no. 12, 2011. [CrossRef] [Google Scholar] [Publisher Link]
[4] Katja Kircher, and Christer Ahlstrom, “Predicting Visual Distraction Using Driving Performance Data,” Annals of Advances in
Automotive Medicine/Annual Scientific Conference, vol. 54, pp. 1-10, 2010. [Google Scholar] [Publisher Link]
[5] Yulan Liang, Michelle L. Reyes, and John D. Lee, “Real-Time Detection of Driver Cognitive Distraction Using Support Vector
Machines,” IEEE Transactions on Intelligent Transportation Systems, vol. 8, no. 2, pp. 340-350, 2007. [CrossRef] [Google Scholar]
[Publisher Link]
[6] Yulan Liang, John D. Lee, and Michelle L. Reyes, “Nonintrusive Detection of Driver Cognitive Distraction in Real Time Using
Bayesian Networks,” Transportation Research Record: Journal of the Transportation Research Board, vol. 2018, no. 1, pp. 1-8, 2007.
[CrossRef] [Google Scholar] [Publisher Link]
THANK YOU