Drowsiness Detection using Contactless Heart Rate Monitoring and Eye Tracking

Detecting drowsiness has many important applications, particularly in driver safety. Blink duration and heart rate variability are two important metrics that can be used to determine how drowsy someone is, and can be determined using image processing techniques. In order to determine blink duration, training images of open and closed eyes are used to generate a fisherimage using Fisher linear discriminant analysis. This fisherimage can then be used to determine whether the eyes are open or closed at a given time, allowing for determination of blink duration. In order to determine heart rate variability, two methods were used to determine the blood volume pulse, independent component analysis and chrominance based. Eye detection yielded very good results, with 71% to 97% accuracy in classifiying open/closed eyes. Heart rate estimates were a little less accurate, with a mean error of roughly 16 BPM for the ICA method and 13 BPM for the chrominance based method. Overall, these results show that detecting drowsiness using simple RGB cameras is very promising.