What ifthe light in the room could sense you waving your hand as you enter? And whatif it responded by introducing minute light changes that instructed your smartcoffee machine to switch on? Researchers at Dartmouth College have developed asensing system called LiSense that aims to make the light around us"smart." Not only does it use light to sense people’s movements, butit also allows them to control devices in their environment with simplegestures, using light to transmit information.
The goal is to use lightto gesture and interact with objects in a room, just like how you’d use a Wiior Kinect to interact with a TV. Xia Zhou, lead author of a paper on theresearch, wants to use smart light to turn every indoor space into a cognitivespace.
"Using purelyvisible light, we can not only stay connected to the internet, but also havethe environment know and respond to what we do, how we behave, and how wefeel," Zhou tells Gizmag. "Smart light can bring intelligence to allthe devices immersed in the light and allow them to act based on ourbehaviours."
To get LiSense to track aperson’s movements only through light, the researchers built a light-sensingtestbed, with LED lights in the ceiling and light sensors on the floor. Thesystem uses the shadows cast by a person standing on the testbed to reconstructtheir 3D human skeletal posture in real time. LiSense, the team states,essentially works on the same principle as a shadow puppet, where a hand heldbefore a light blocks certain light rays and not others.
"Consider a personstanding under several lights," Zhou explains. "If we can recover theshadow cast by each light in a different direction, we can aggregate the shadowinformation and collect the blockage information of a large number of light rays.We then use the information to search for a 3D skeleton posture that bestmatches the blockage information revealed by these shadows."
To gettheir shadow-based human sensing to work, the researchers had to crack twocritical challenges. Since multiple ceiling lights lead to diminished andcomplex shadow patterns on the floor, they came up with light beacons. Theseseparate the light rays from different light sources to recover the shadowpattern cast by each light. The team also designed an algorithm capable oftaking the collected low resolution, 2D shadow maps from sensors in the floorand reconstructing a person’s posture in 3D.
LiSense makes use ofadvances in visible light communication (VLC), where information is encoded aslight intensity changes at high frequency. Since most of the smart devicescommonly available contain light sensors, they’re capable of receiving data bymonitoring changes in light.
"The light changesare imperceptible to human eyes, yet the light sensors on our smart devices cansense the changes and decode data," Zhou tells us.
In tests, LiSense wasable to reconstruct a 3D user skeleton within 16 milliseconds (ms) in realtime. It was also able to produce shadow maps of all the LEDs every 11.8 ms,which is comparable to capturing video frames (without using any cameras). Thesystem was also found to be robust in various light settings with users ofvarying body sizes and shapes.
Going forward, theresearchers plan to minimize the number of light sensors used and have themseamlessly integrate into the environment. They also want to make the systemsense more than low-level gestures, which could potentially open up all kindsof novel applications. For instance, smart light could monitor both our healthand behaviour, to catch diseases in their early stages.
"If the light aroundus continuously monitors how we move and gesture over time, it might helpdetect early symptoms of diseases such as Parkinson’s, which hasmovement-related symptoms," Zhou explains. "Right now patients haveto wear and carry bulky, cumbersome devices, or doctors have to videotape them.Light provides a new possibility—no on-body devices, no cameras."
A light-based sensingsystem also has numerous advantages over existing technologies, the team says.There’s no electromagnetic interference, it doesn’t penetrate walls and itreuses existing lighting infrastructure. It’s also secure, has a bandwidth10,000 times greater than the radio frequency spectrum, and isn’t limited toclassifying a pre-defined set of gestures and activities.