A Computer That Notes Your Figure

Computer imaging and digital photography have come a long way in the past few years, but it's still a little flat; cameras have no depth perception and directly record only two dimensions. But that changed Monday, when computer-vision startup Canesta Inc. unveiled what it says is a new, inexpensive way for computers to sense the world around them in all its depth.

The company is touting "new electronic-perception technology that has the capability of constructing 3-D images of its nearby surroundings in real time," says James Spare, Canesta's VP of product marketing. A regular digital camera has an array of sensors that record brightness from ambient light reflected onto them, Spare says, thus obtaining "flat" images. But devices using Canesta's sensor and software emit their own infrared light, measure how long it takes to bounce back to the sensors, and do real-time processing to figure out how far away objects are from the camera--a lot like how radar works. "What that creates is a contour map," Spare says.

The sensors could watch someone's finger movements or track the movement of a pen and convert them into keystrokes, for example. Security systems could recognize people by the contours of their bodies and faces. Video games could even watch a player's body and translate kicks and punches into a digital character.

Spare says the first "perception chips" with embedded software should be available to manufacturers in the third quarter of the year, at prices comparable to regular digital-camera chips. The first commercial products--probably part of input devices for mobile phones and wireless devices, Spare says --should be on shelves in the first half of 2003.