Microsoft Research has demonstrated a prototype keyboard that can be used to recognize gestures. The Type-Hover-Swipe
keyboard is able to detect fingers and entire hands hovering just above the keys, allowing users to perform various maneuvers typically used on a touchscreen or a tablet but without having to make their hands travel far from the home position.
The keyboard uses a collection of infrared proximity sensors placed strategically between the rows of keys, connected to a circuit board inside the keyboard membrane. From the 64 pixels of data, the researchers reworked a machine learning algorithm used for static classification to support gestures, giving the system an average per-frame classification accuracy of 75.6 percent, or 89.9 percent in half-test, half-training cross-validation.
While it could be considered similar to the company's own Kinect
motion tracking system for the Xbox and the Leap Motion
controller, though comparatively low-resolution, the prototype does offer a few advantages. The system is contained within an existing computer component already widely used, instead of requiring an extra device to be placed on the desk, and the shallow range of detection above the keyboard will help avoid false positives that can crop up with the other two systems.
The Type-Hover-Swipe keyboard is still in development, and is highly unlikely to be released in its current form.