Researchers at the University of Cambridge have developed a robotic sensor that uses artificial intelligence to read braille at twice the speed of human readers. The machine learning algorithms help the robot read braille at 315 words per minute with close to 90% accuracy. The device was not built as assistive technology, but it can help in the development of highly sensitive robot hands or prosthetics.
Human fingertips are extremely sensitive and provide important information about the world. They can detect features of materials or help gauge the amount of force needed to hold an object. The researchers, led by Professor Fumiya Iida, are working on developing robot hands that can match the sensitivity of human fingers. This is a challenging task due to the softness of human fingertips and the need for a lot of sensor information.
The researchers used an off-the-shelf sensor to develop a robotic braille reader. This sensor has a camera in its ‘fingertip’ and uses a combination of information from the camera and the sensors to read. The team developed machine learning algorithms to help the robotic reader ‘deblur’ the images before recognizing the letters. Despite the fake blur used to train the algorithm, the reader was surprisingly accurate at reading braille.
The researchers plan to scale the technology to the size of a humanoid hand or skin. The study was supported in part by the Samsung Global Research Outreach Program. This robotic sensor is an important step forward in the development of highly sensitive robotic hands and prosthetics that can be used for various applications beyond just braille reading.