Could a low-power chip used for processing 3D camera data help the visually impaired find their way around?
The chip, developed by researchers at the Massachusetts Institute of Technology, consumes a mere one-thousandth of the power of conventional computer processors executing the same algorithm. It powers a complete navigation system’s prototype with a binoculars-case size that can be carried around the neck.
Developed at MIT's Computer Science and Artificial Intelligence Laboratory, the mechanical Braille interface carried by the user conveys data on the distance to the closest obstacle in the user’s direction.
The team addressed the bulkiness of the systems because of the various processing required.
“We wanted to miniaturize this system and realized that it is critical to make a very tiny chip that saves power but still provides enough computational power,” said first author Dongsuk Jeon in a statement.
The algorithm involved converting data from the 3D camera into helpful navigation aids. As any 3D camera can have its output converted into 3D representational tool “point cloud,” the group’s algorithm clustered together the resulting points to pinpoint flat surfaces and measured the unimpeded walking distance of different directions.
The team further tweaked this algorithm to reduce unnecessary power consumption, such as the common energy drain caused by fetching data from the main memory.
While already a far cry from its bulky, obtrusive predecessors, the size of the prototype navigation system can be reduced, added the researchers.
The team reported these developments – which could serve as a potential high-tech “guide dog” for the blind – in a paper showcased this week in San Francisco at the International Solid-State Circuits Conference.
The work was funded in part by the Andrea Bocelli Foundation established by the blind Italian tenor Andrea Bocelli.
Technology has other potential aids and surprises in store for the visually impaired, including the promise of a Braille tablet device that would allow users to read straight off the screen with the aid of so-called microfluidic bubbles.
Photo: Roland DG Mid Europe Italia | Flickr