Drowsiness Detection and Anti-Sleep System for Drivers
Jonathan Dulce & Harold Johnson
Utilising effective, real-time and non-intrusive methods to detect drowsiness on the road.
This project pairs the use of ballistocardiography (BCG) sensors within the driver’s seat to monitor driver heart rate variability, as well as the use of computer vision to track the driver’s face behaviour for indications of drowsiness. Unlike other methods of heart rate monitoring, BCG provides a non-intrusive platform for detecting heart rate variability that can indicate drowsiness, without distracting the driver. Combined with an IR camera for tracking driver face behaviour of the eyes, mouth, and head tilt, this system can alert the driver if they are drowsy or fatigued in real-time and can potentially prevent accidents.
Page Views:
The prevalence of drowsy driving and casualties as a result of drowsy driving remains alarmingly high - with it being a primary factor in approximately 20% of all fatal accidents. A direct drowsiness detection system which can monitor the driver’s vitals and behavioural changes can provide a cost-effective, affordable, and efficient means of improving road safety.
This project pairs the use of ballistocardiography (BCG) sensors within the driver’s seat to monitor driver heart rate variability, as well as the use of computer vision to track the driver’s face behaviour for indications of drowsiness. These peripheral sensors are controlled and have their data processed with a Raspberry Pi, with the camera positioned in front of the driver on the dash, and the BCG sensors positioned on the seat underneath the driver’s lower thigh to record their heart rate.
Piezoelectric film sensors were used for BCG. BCG peaks are detected, extracted, and processed based on their high and low frequency components to track heart rate variability.
An IR camera is used to conduct computer vision so that this system can perform at night. Key face landmarks are extracted and used to monitor facial behaviours such as eye aspect ratio, percentage of eye closure, blinking rate, yawning, and head angle to detect drowsiness.
Due to COVID limitations, development, testing and verification of our system was performed on the Bed-Based Ballistocardiography database and NTHU database.
