Placebo : an interactive notice board system

Abstract

The system proposed in this thesis acts as a prototype of a notice board system that detects who is standing in front of it through the use of a webcam, if the person is recognised personalised information is displayed to the user. The prototype developed focuses on the particular use of the notice board to display a faculty's timetable. When a student from that faculty is recognised, their classes are highlighted on the timetable displayed by the notice board. The thesis explores a number of different solutions for face recognition and attempts to find a more efficient way of storing face data gathered to be used during the recognition stage. Other problems tackled include dealing with spoofing (tricking the system into authorising a user as someone else) and what to do when more than one person is in front of the camera. The end system manages to recognise and deliver personalised results successfully. The anti-spoofing measure implemented is a blink detection stage which initializes the recognition stage. Results show that attempts to fool the system by holding up a picture of a person that should be recognized were completely ineffective and that the system can't be spoofed this way. The system also manages to save on average 75% of space required to store the images of subjects to be recognised by only storing the important features extracted from the image. This was achieved without any loss to the performance of the recognition engine. Though in theory the prototype can handle more than one user at a time, in practice the load of tracking more than one face proved extremely computationally intensive and results in this respect were poor.