Driver Drowsiness Detection and Warning using Facial Features and Hand Gestures

According to National Highway Traffic Safety Administration (NHTSA), drowsy driving is one of the primary causes of accidents. Numerous valuable lives can be saved, accidents can be reduced or avoided, and the cost of injury and damage to infrastructure may be reduced with a timely alert or warning to the negligent driver. Advanced Driver Assistance Systems (ADAS) consists of the active safety system which includes the detection of the driver's face to determine their level of drowsiness. This paper provides a camera-based technique which relies on fiducial components, such as lips, eye movement, and hand gestures of the driver which are often natural responses of a human to yawning. A front camera installed on the windscreen is used to continually monitor the driver and Raspberry Pi is utilized for processing the images. The proposed warning system gives an audio warning when the driver is yawning or going into the state of drowsiness. Results illustrate that the proposed technique is effective at detecting signs of driver's drowsiness and yawning. It can differentiate between when the driver's hand is placed over the mouth to infer it as yawning and when it is touching other parts of the face to infer it as not yawning.