This Content Is Only For Subscribers
Health, UK (Commonwealth Union) – A team of researchers at the University of Cambridge has designed a robotic sensor enhanced with artificial intelligence capabilities, enabling it to interpret braille at speeds approximately twice as fast as the average human reader. Through the implementation of machine learning algorithms, the robotic sensor swiftly traverses lines of braille text, achieving a reading speed of 315 words per minute with nearly 90% accuracy.
Although the primary goal of the robot braille reader was not centered around assistive technology, the researchers highlight its exceptional sensitivity in reading braille as a valuable characteristic for testing and advancing the development of robot hands or prosthetics that mimic the sensitivity of human fingertips. The findings are detailed in the IEEE Robotics and Automation Letters journal.
The sensitivity of human fingertips plays a crucial role in gathering information about our surroundings. It allows us to discern subtle changes in material textures and determine the appropriate amount of force needed to grasp objects, such as delicately picking up an egg without breaking it or securely holding a bowling ball, as indicated by researchers of the study.
Efficiently replicating this high level of sensitivity in a robotic hand poses a significant engineering challenge. Professor Fumiya Iida’s research team in the University of Cambridge, Department of Engineering is actively working on developing solutions for this challenge and addressing other tasks that humans find effortless but prove challenging for robots.
“The softness of human fingertips is one of the reasons we’re able to grip things with the right amount of pressure,” explained Parth Potdar from the University of Cambridge, Department of Engineering and an undergraduate at Pembroke College, the 1st author of the paper. “For robotics, softness is a useful characteristic, but you also need lots of sensor information, and it’s tricky to have both at once, especially when dealing with flexible or deformable surfaces.”
The utilization of Braille as a testing ground for a robot’s ‘fingertip’ is ideal due to the high sensitivity required to read it. This is because the dots in each representative letter pattern are closely spaced. The researchers employed an off-the-shelf sensor to create a robotic Braille reader that closely mimics human reading behavior.
Co-author David Hardman, from the Department of Engineering, pointed out the limitations of existing robotic Braille readers, indicating that presently there are existing robotic Braille readers, but indicated that they only read one letter at a time, which is not how humans read. He further explained that these readers function statically—touching one letter pattern, reading it, lifting from the surface, moving over, lowering onto the next letter pattern, and so forth. The goal is to develop a system that is more realistic and significantly more efficient than current approaches.
The researchers employed a robotic sensor equipped with a camera in its ‘fingertip’ for reading, utilizing a combination of information from the camera and sensors. Addressing the challenge of motion blur, Potdar indicated that it is a difficult problem for roboticists as there is plenty of image processing that are required to be done to take away motion blur, which is time as well as energy-consuming.
To tackle this, the team developed machine learning algorithms that allowed the robotic reader to ‘deblur’ images before recognizing the letters. The algorithm was trained on a set of sharp images of Braille with artificially applied blur. Once the algorithm learned to deblur the letters, a computer vision model was employed to detect and classify each character.
After the incorporation of these algorithms, the researchers tested the reader by swiftly sliding it along rows of Braille characters. The robotic Braille reader demonstrated an impressive reading speed of 315 words per minute at 87% accuracy, surpassing human Braille readers in both speed and accuracy, performing at twice the speed while maintaining comparable precision.
