Lenovo Face Recognition

Some of our Lenovo notebooks come with face recognition software, which is actually a reemergence of an old idea. We had facial recognition capabilities a long time ago. My first experience was with a ThinkPad T23 with an external camera mounted on the system's UltraPort. It didn't work well at all. Most likely because the cameras of that era were terrible. I think they were 0.3 megapixels, but may not have even been that much. Now that some systems include integrated cameras with much better quality (1.3MP), facial recognition has become much better. The included software lets you log onto your Windows account simply by sitting in front of your system. Your face is your password. What is much cooler is that it is very user friendly for multiple user accounts. For example, let's say you have three Windows accounts – Mom, Dad, and Sis. If you have associated their faces with their respective user accounts, the system determines which person is in front of the computer when Windows boots and automatically logs them onto the right account. In practice this works very well and is extremely fast at recognition. I was able to test this with several of my colleagues and each time all they had to do was sit in front of the computer and the system took care of the rest. Depending on the software used, face recognition uses multiple techniques to identify a person's face. Some of the more advanced programs use texture mapping in which a person's skin texture is analyzed and matched. Most however, define nodal points on a person's face and then use software to mathematically represent those points. Things measured include distance between the eyes, width of the nose, length of the jaw line, or shape of the cheekbones. Together these concatenate a numerical code which is stored in a database for later retrieval. One particular aspect of the software Lenovo uses is rather freaky. When you sit down in front of the camera, the system generates two white dots that follow your eyes. Of course, this is completely harmless and is nothing more than a few white pixels shown on screen. However, when I see this, I immediately think that there are two lasers drilling holes into my corneas. Neurotic, yes, but that doesn't stop me from wishing there was a way to turn this off. Others I polled liked the feature, so your experience may be totally different. Of course, a feature like face recognition invites play, and what better way to play than to try and fool the software. First up was an 8 x 10 color glossy photograph of yours truly (with circles and arrows and a paragraph on the back). No matter how I held the photograph, no matter whether the security settings were set high or at their lowest setting, no matter what angle I held the photo, I was not able to use it to log onto the system. The result was exactly what I had expected – that the software was smart enough to distinguish a face from a picture of a face. My next experiment was to see if the system recognized all types of faces. Sitting at home, I tried to enroll both of my birds. The software wouldn't enroll them. I thought that perhaps that since their eyes were not at the front of their heads, the camera couldn't see both at once and therefore couldn't get a good reading. So next I tried my friend Jim's cat. Same result – the software refused to accept the cat, but did happen to enroll Jim's face when he got too close to the camera while holding the cat. Just for good measure, I tried another friend's dog. On all of these, the software wouldn't enroll the animals. So it seems that only human faces are recognized and accepted by the software. It was time for one last experiment. Many people are familiar with the fabled two-key system to launch nuclear missiles. In this scenario, no one person can launch the weapons. It takes two people standing several yards apart turning the keys simultaneously to make it work. What if we could apply that same principle here – requiring TWO faces, not one, to log onto a user's account? While we wouldn't be launching nuclear weapons, such a scenario could be useful in any number of ways. So I sat side by side with my friend Aimee in Lenovo's Product Reviews lab to see if we could make this work. We created a new Windows account and tried to enroll both of our faces at once. The lighting was good, and since we were sitting at the same height right next to each other, we were more or less equal in the eyes of the camera. The software indicated a good capture, so we logged off. We sat down together and the system logged us on successfully. Then we tried just sitting in front of the system one by one. It logged on Aimee, but didn't accept my face. For good measure, we tried the whole setup one more time. Again, it picked Aimee's face over mine. So not only is the software biased towards human faces, it also is biased towards women's faces. I guess I should feel marginalized, but I'm more amused than anything else. Overall I was impressed with the software. It was fast, accurate, and easy to use. Supposedly there is a password management feature where you can use your face to act as your password – much like on our fingerprint reader models. I haven't tried it out yet, but if it is just as easy to use, then I think it would be a very viable solution.