The purpose of a college project is learn stuff and putting what you have learned to practical use. Matlab is a bad choice if you want to learn stuff as it makes everything too easy with its tool kit's and you will never learn the how and why of any thing you implement. I can put together a basic face recognition program in Matlab in under an under an hour and that's how simple it makes it.
Mablab is a tool that is intended for advanced research work by people who already know/understand how the basic stuff works. For Academic work, Personally I would not recommend use of Matlab for anything other than a PhD thesis or for complicated simulations that need to be done in a short span of time.
I have been interviewing candidates for my company and although they have mighty sounding projects, they were all done in Matlab and they were totally clueless about the how and why of their projects. So I would recommend a different programming environment like VB, C# or Java if you want to learn about the actual subject you are interested in (I hope you are really interested in Face Recognition and not just interested in handing in some mighty sounding project) while at the same time not spending too much time on UI or other low priority programming details.
Next coming to the topic of your project, Face Recognition, I will not give you source code, but I will give you a clue of where to start. There are numerous Pattern Matching/Soft computing techniques/tools for doing face recognition. There's Principal Component Analysis, Neural Networks, Wavelet Analysis and others combination of one or more of these tools/techniques and each have their merits and accuracy of recognition. Every technique involves extraction of a set of features (called feature set or feature vector) from sample face photos and using that for pattern matching.
You can start by reading and understanding the theory. Next you can download a face database. Search in Google with key words face database. Most of these have a set of faces taken from multiple subjects. For example, the one I used at college had 10 photos each of 40 subjects for a total of 400 photos. You can use 5~7 photos for creating the reference feature sets and then use the rest for testing.