Forearm Gestures Remotely Control Computers and Drones

The MYO armband translates electrical activity produced by muscles into commands for gadgets

I expected “gesture control” to be immediately intuitive. But as I slip on the MYO—a flexible band that fits around my forearm—a cursor on a laptop in front of me begins somersaulting wildly across the screen, tracking my erratic arm movements. I focus, slow down and try to get a feel for this new tool.

“Move your wrist right—and now left,” instructs Stephen Lake, co-founder of Thalmic Labs of Waterloo, Ontario, the start-up behind the MYO (named after a biological prefix denoting muscle). As I do, the engineering interface on Lake’s computer screen registers a burst of raw data—peaks and dips of scrolling electrical activity produced by my engaged skeletal muscles. Then, the program flashes the words “right” and “left,” confirming that it understood my actions. I’m beginning to get the hang of this.

I’m in a New York City office, where Lake is offering the first hands-on demonstration of the MYO, a new gesture control interface. The armband has insulated electrodes that detect small volts of electricity that muscles produce when they expand, contract or move in any direction. The band transmits those data wirelessly to software, which translates them into commands for a computer, drone or other electronic device. The idea is to control these devices hands-free, and without the need for cameras that would track my motions.

The MYO prototype resembles a clunky bracelet of the type Wilma Flintstone might wear. The final product—the first batch ships out at the end of the year—will resemble a sweatband, Lake says. The prototype is fashioned out of 3-D–printed black plastic, embedded with several muscle activity sensors. They act as electromyographs, or instruments that detect minute electrical signals on the order of microvolts, produced by activated muscles. “The challenge is picking up those tiny muscle activity signals and ignoring all the noise,” Lake says.

An inertial sensor, embedded in one of the MYO’s segments, registers motion made with the arm, such as a rolling wave or a back-and-forth swing. Using a large set of data, Lake and his co-founders applied machine learning to train the MYO to recognize specific signals while canceling out background noise. “What I am impressed with about the MYO is the combination of state-of-the-art pattern recognition and machine learning algorithms to detect gestures, with a strong base of acquiring data,” says Daniel Stashuk, an electrical engineer at the University of Waterloo who has no financial ties to Thalmic Labs, in a phone interview. “Marrying those two things together is quite useful.”

So far, the sensors can recognize around 20 gestures, from a sweeping arm to a clenching fist. On the finest end of the spectrum, the MYO responds to a thumb and finger pinching together. “It’s not that we couldn’t detect smaller motions, but if we did, there would be so many false positives,” Lake explains.

See Also

Smartphone understands gestures

The MYO’s greatest limitation, Lake thinks, is the fact that the user must wear it. If the armband is not wrapped around the forearm, it cannot detect movement. Current challenges to improving the MYO’s performance, he adds, include better defining an intuitive set of gestures that could be applied across a wide variety of applications. The team is also working to refine the MYO’s algorithms to improve balance between sensitivity and false gesture detection.

Read more . . .