Abstract:
Image processing is commonly used to recognize objects. It can be used to
detect human facial expressions and emotions by observing facial features. Capturing
emotions depends on many tools and conditions such as brightness, camera, changing
rate of facial features etc. Emotion recognition is software that allows program to “read”
the emotions on the human face using advanced image processing. In order to understand
not only what a person’s face or image looks like, but also how it looks. Emotion
recognition has been applied in many fields such as Medical, Security, and Business etc.
There are many difficulties in developing a good emotion recognition system for the
human face in real time. Since most of the time facial features of expression and the style
of showing emotion to the outside world is different from person to person. Therefore, it
is very difficult to build an accurate system for real time emotion recognition. This
project is to detect human facial expressions to predict the current emotional state. The
system specially focused on reducing fatal road accidents due to drivers' state of emotion.
Initially it is built to recognize the human emotion through facial expressions and then
evaluated to detect drowsiness using facial landmarks to ensure the safety of the driver.
Training has been done with Kaggle dataset for seven emotional states (Neutral, Happy,
Angry, Sad, Scared, Surprised and Disgust) called universal emotions. In order to predict
drowsiness, it uses specific 12 points on face (6 points on each eye) in shape predictor
68 face landmarks. Evaluated system has given 66% accuracy in testing and drowsiness
alert also showed a very good success rate.
