The Orpheum Children’s Science Museum, 346 N. Neil St., hosted Brain Awareness Day Sunday, which featured exhibits about the brain and how it functions. This event, sponsored by the University’s Neuroscience Program, gave children and their parents a glimpse into how their own minds work through displays and robotics.

One of the exhibits, called “A Camera with a Brain has its Eye on You,” featured a freestanding, self-aiming video camera. The camera uses motion detection, infrared and a face-detection computer program to pick out the most active person in a room and pivots to keep the targeted person within its field of vision.

Thomas Anastasio, professor of neurological sciences at the University, helped develop the camera and has always been interested in how the human brain works. Anastasio focused his studies on how the brain applies information that is gathered by the senses. However, it wasn’t until he was approached by Thomas Huang, professor of electrical and computer engineering, that this interest led him to develop a new technological advancement.

“I was interested in combining brain science and engineering to make a product, and (Huang) said, ‘How about a self-aiming camera?'” Anastasio said.

Anastasio said he and Huang hadn’t considered the application of such technology until they started working with Sylvian Ray, professor of computer science. Ray imagined that this kind of technology could be used in a remote learning situation.

“If a student had a question, the camera would immediately recognize that, hey, this person over here is asking the question, and then the professor could look at the person asking the question on a screen at his location,” Anastasio said.

However, now that the camera has been developed, Anastasio said he thinks the technology would be more likely adapted into a surveillance camera system.

The camera, which is five years in the making, is based on mathematical models of the tectum, a region of the brain that is found in all vertebrates. In mammals, this part of the brain is called the superior colliculus, and it controls subconscious reflex-like behavior in response to things in the environment.

“For example, if a fly buzzes by your ear, you will turn around and look at it before you even think about what you’re doing,” Anastasio said. Using Anastasio’s models of how the tectum applies information, Ray created a software program to make the camera locate and follow people in an environment. Until recently, however, the camera was not very accurate, Anastasio said.

Initially, Anastasio and Ray’s camera had several limitations. For one, it could only move in horizontal directions. This prevented the camera from closely following a person moving nearer or farther away from the camera.

Also, it only used visual and audio information to detect movement in a room.

With the addition of the infrared detector, the camera’s effectiveness is not hindered by the amount of light available or the acoustics in a room. Instead of relying on sound, the camera could identify the heat emitted from a living person, thereby making the camera ideal for surveillance applications.

“What we did was we made warm, moving things especially interesting to the camera,” Anastasio said. “We also added this fancy detection system that recognizes human faces.”

The detection system Anastasio referred to was actually developed in 2001 by Paul Viola and Michael Jones, two Compaq researchers at the Cambridge Research Laboratory in Massachusetts.

The program, called a Robust Real-Time Object Detection system, uses shadows on an image to determine if the object might be a face. The program is designed to recognize dark, shadowy areas that are neighboring brighter areas of the same size. For example, the eyes are a darker portion of the face than the cheekbones and nose, which lie directly beneath them. Also, each eye is darker than the bridge of the nose found between them.

Viola and Jones then created a computer software program that would look for objects that had both of those contrasting conditions in the same relative area.

“As it turns out, only faces and very few textures have these characteristics,” Viola said in e-mail.

The system works so well that it detects 100 percent of all faces within the camera’s view, and it mistakes less than 10 percent of non-faces for faces. However, some of the 10 percent includes models of people.

“It will sometimes confuse statues (of people),” Viola said, “but these are really faces, even if they’re inanimate.”

Despite the high accuracy of Viola’s program, he stresses that it does not actually recognize faces; it only recognizes patterns of shadows that are more face-like than others.

“My system does not do person recognition,” he said. “You cannot tell if George Bush or Dick Cheney is standing in front of the camera – just a person.”

Another limitation of Viola’s program is that the person must be facing the camera in order for the program to detect the shadows on the eyes – tipping the camera off that there is a face there.

“With our camera,” Anastasio said, “a profile image of a person is still visible because of the motion detectors and infrared cameras.”

In addition to the new technology found on Anastasio’s self-aiming camera, the camera, which is small enough to fit on top of a classroom desk, can now rotate vertically, as well as horizontally and all of the directions in between.

Anastasio believes that with these new improvements, his invention would be ideal for surveillance systems or even research about brain functions.

“This little machine could be used to test ideas about memory or image selection or cognition,” Anastasio said. “So the robot, which was designed on the basis of our ideas about the brain, could actually help us test new ideas about brain function.”