PROBING FACIAL GESTURES FOR DETECTING DROWSINESS

Project Title

PROBING FACIAL GESTURES FOR DETECTING DROWSINESS

Project Area of Specialization Electrical/Electronic EngineeringProject Summary

The basic idea behind such a project is to create a device that can identify any human fatigue and can provide a timely alert. Drivers who do not take regular breaks when driving at night run a high risk of becoming drowsy, a situation that they might fail to acknowledge early enough. Driver mistakes and negligence are leading to traffic accidents these days, the greatest driver mistake is induced by the driver’s drowsiness, fatigue, and aggressive behavior.  According to reports published by the Pakistan Bureau of Statistics Government of Pakistan, there is about 11,121 number of accidents have occurred from 2017-18 and nearly 5,948 people killed as well as 13,134 vehicles are involved as a result.

Research studies also suggest that early detection of drowsiness. Currently, several techniques are being used for drowsiness detection like EEG Based Detection, Artificial Neural Networks, Vehicle-Based approach, and physiological based approach but all these techniques have some disadvantages.  For example, in the Artificial Neutral Networks Technique Several neurons are used to recognize driver drowsiness, but only one neuron is not very reliable, and the performance is not good compared to even more than one neuron while using EEG Based Detection, drivers will wear electrode helmet while driving. This helmet is fitted with multiple electrode sensors that are mounted at the right place and get brain data. So, these types of techniques are much costly as compared to the Behavioral-based approach.

To increase the chances of early detection, this project is geared towards drowsiness detection in drivers using a Behavioral-based approach which is a technique that is not only extremely precise but is also cost-efficient. In this technique a person's eye blinking frequency, head position, etc. is tracked via a camera and the person is notified if any of these indications of drowsiness are observed. The results of drowsiness detection are the innovative step for drivers and many other's lives.

Project Objectives

Nowadays, driver safety is one of the most wanted systems to avoid accidents. The main objective of this project is to create a system for drowsiness detection which alert driver’s whenever they fall into the drowsy state.  

1. Our project aim is to enhance the safety system. To improve safety, we detect the driver's eye blinks and predict the driver's status.
2. Our proposed method is to design and develop a low-cost system, which is based on computer vision platform for drowsiness detection.
3. The objective is to overcome the problem related to the accidents related to drivers experiencing fatigue leads to need arises to design a system that keeps the driver focused on the road.

Project Implementation Method

Python and OpenCV implemented drowsiness detection which includes the following steps:

• Successful runtime capturing of video with the camera
• The captured video was divided into frames and each frame was analyzed 
• Successful detection of the face followed by detection of the eye

If closure of the eye for successive frames were detected, then it is classified as a drowsy condition else it is regarded as normal blink and the loop of capturing an image and analyzing the state of the driver is carried out again and again. Using the EAR variable to predict the driver’s drowsiness level. The eye closure rate is a behavioral indication for drowsiness detection. After the detailed study of behavioral measurements to detect drowsiness, we selected the eye blinking duration of the driver as the behavioral measurement to detect drowsiness.  To calculate the blink duration, the first thing we did face detection.  To do the face detection we use the “Harr cascade Algorithm” and to detect eye in the face we use the “Facial landmarks detector”.

Here, behavioral measurements are proposed for drowsiness detection.

1. At this time in Pakistan no such type of system that detects the driver drowsiness so it will be the first prototype that will detect driver drowsiness and enhance the safety system.
2. We can also provide the user with an Android application that will provide his / her drowsiness level information during every ride. The user would know Normal State, Sleepy State, how many times the eyes have blinked based on the number of frames captured.
3. Easy implementation due to readily available hardware and software.
4. Efficient system to identify user attentiveness based on fatigue detection.

Basic steps of the behavioral measures are as follows,

• Study on behavioral measures used to detect drowsiness
• Selecting facial landmarks analysis to detect drowsiness
• Analyzing Left eye and Right eye ratio for calculations of the EAR (Eye Aspect Ratio)
• Selecting a suitable range of EAR for the implementation

Here, in the first figure, we detect face and eye using Haar cascade classifier from the static picture. And in the second figure, we apply a facial landmark detector on video streaming, and we detect the eye, mouth, nose coordinates. 

Figure 1: Face and eye detection               Figure 2: Facial landmarks detection

Benefits of the Project

1. At this time in Pakistan no such type of system that detects the driver drowsiness so it will be the first prototype that will detect driver drowsiness and enhance the safety system.
2. We can also provide the user with an Android application that will provide his / her drowsiness level information during every ride. The user would know Normal State, Sleepy State, how many times the eyes have blinked based on the number of frames captured.
3. Easy implementation due to readily available hardware and software.
4. Efficient system to identify user attentiveness based on fatigue detection.

Technical Details of Final Deliverable

1. By using camera, we detect the driver drowsiness by this we are detecting face, eye and yawing.
2. Camera is attaching with the raspberry pi which is work as a processor and providing some memory for results.
3. We use a buzzer which is used to alert driver whenever he falls in drowsy state.

Final Deliverable of the Project Hardware SystemCore Industry TransportationOther Industries Security Core Technology Internet of Things (IoT)Other Technologies Wearables and Implantables, OthersSustainable Development Goals Reduced Inequality, Sustainable Cities and CommunitiesRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	26900
Raspberry Pi 4 (4GB)	Equipment	1	15000	15000
Camera	Equipment	1	3000	3000
Buzzer	Equipment	1	200	200
Memory Card	Equipment	1	1000	1000
Power Cable	Equipment	1	200	200
Supply Charger	Equipment	1	1000	1000
LED	Equipment	3	100	300
Acrylic Sheet	Miscellaneous	2	500	1000
Laser Cutting	Miscellaneous	1	2000	2000
Measuring Tape	Miscellaneous	1	1000	1000
Scale	Miscellaneous	1	100	100
Printing	Miscellaneous	600	3	1800
Double Tape	Miscellaneous	1	300	300