I created a security robot running onboard facial recognition using the Parallax EDDIE robot platform. The idea was that the guard robot would be able to tell the difference between potential theives versus friends and family. The facial recognition technique works by taking a set of different faces, each of which is a data vector, and computing the eigenvectors of the set, also called "eigenfaces". A particular individual's face can be represented as a weighted linear combination of eigenfaces. The weights for a given individual are unique as they specify the importance of the different principal components, which are in practice the common human facial features and possible lighting conditions, gender differences, etc. Given an unknown query face, we can solve for the weights we'd need to reconstruct the query face from our eigenfaces. This set of weights is compared to our database of weight sets for known individuals to classify the query face. I was named as one of ten national finalists by Microsoft for my project, one of only two students to be finalists in the competition. Please see my video and old blog below for details on the project.