(Currently on hold)
The goal of this project was to identify if a person was lying by analysing a video feed and extraccting facial microexpressions. We used OpenCV's harrcascade classifiers for face detection and eye detection. This ROI was then amplified using Eulerian video magnification for visualizing the human pulse. This algorithm amplifies the periodic color variation in the human face, using which we can obtain the pulse rate. Now this can be one of the features contributing to the classification.
We couldn't think of other distinguishing features that could be used for classification (hence on hold).