An optical sensor that can recognize characters and words written in Braille could make information more accessible to people with a visual impairment. [Image: Zhuo Wang, Beijing Normal University]
Researchers in China have demonstrated an optical sensing system that can read Braille from the pressure between the raised dots and a flexible probe that mimics a person's finger (Opt. Express, doi: 10.1364/OE.546873). The tactile device, which is made by embedding a fiber-based sensor into a soft material similar to human skin, could pave the way for automated readers that convert Braille into text or speech. Such smart translation solutions would help to make information more accessible for people with a visual impairment, particularly for those who lose their vision later in life and have no experience of reading Braille.
Leveraging phase change
Optical sensors have been explored as a precise and flexible technology for recognizing the dotted patterns, but existing devices lack the sensitivity to decode smaller dots or those with any imperfections. The high-precision sensor demonstrated by the researchers exploits an optical ring resonator, in which the light from a tunable laser source circulates around a continuous loop of fiber. Embedding this optical device within a skin-like material makes it sensitive to pressure, with the phase of the light changing when the flexible probe touches the Braille dots.
This phase change causes small variations in the resonant frequency of the resonator, which can be detected by a modulation scheme that synchronizes the frequency of the input laser to that of the resonator. To improve accuracy, the researchers also implemented machine-learning methods that use previous experimental data to distinguish between similar signals, making the system more robust to slight variations in the contact pressure.
Fast and accurate
The researchers tested the performance of the sensing system by moving the tactile probe across a sample of Braille, first using a mechanical force tester and then by hand. In both cases, the system recognized eight different Braille patterns with an accuracy of 98.6%, and it correctly read five different words even when there were slight discrepancies in the individual characters. "This system is far more precise than older Braille-reading technologies that might miss imperfectly pressed dots," said team member Rui Min, Beijing Normal University.
The researchers are now working to reduce costs and improve the durability of the device. They are also planning to investigate other machine-learning models that would enable the system to tackle more complex Braille-reading tasks. “This technology could help Braille become more widespread in public spaces, on digital platforms and in education,” said Heng Wang, Shenyang Aerospace University, China.