Flexible Optical Skin Enables Fast and Accurate Braille Reading

Researchers have developed a flexible optical skin capable of reading Braille with speed and accuracy. This technology has the potential to enhance information accessibility for individuals who are blind. The findings were published in the journal Optics Express.

Braille, which is made of tiny raised dots, requires exact sensors to be read correctly. Traditional sensors struggle with this, especially for dynamic tasks like reading Braille as a person moves their finger over it in real-time. Our sensor can handle tiny Braille details accurately and quickly.

Zhuo Wang, Research Team Leader, Beijing Normal University

The sensor consists of an optical fiber ring resonator embedded in soft PDMS material, forming a flexible optical skin. When pressure is applied, the sensor detects frequency changes in the light and converts these variations into readable data.

Testing showed that the system could quickly and accurately interpret Braille letters, numbers, and punctuation when integrated with neural network-based data processing techniques.

This technology could lead to faster, more accurate smart readers that read Braille and convert it to speech or text without the user having to learn Braille. This could help Braille become more widespread in public spaces, on digital platforms, and in education. The impact could also extend to other fields where sensitive tactile sensing is needed, from smart medical devices to the next generation of robots.

Heng Wang, Research Team Member, Shenyang Aerospace University

Automated Braille readers are particularly valuable for older individuals who lose their vision later in life, as learning Braille can be challenging. While optical tactile sensors are known for their accuracy, flexibility, and resistance to interference, their sensitivity to small Braille dots and susceptibility to environmental disturbances remain limitations.

To address these issues, the researchers incorporated an optical fiber ring resonator—a loop of optical fiber that channels light—into the soft PDMS material. This configuration provided the sensor with flexibility similar to human skin, allowing it to bend and detect pressure effectively. When pressed onto Braille dots, the pressure slightly bent the optical fiber, altering its resonant frequency.

The researchers used the Pound-Drever-Hall (PDH) frequency-locking technique to ensure accurate readings by maintaining a stable light signal even in dynamic conditions. Machine learning further improved the sensor's accuracy. A Multilayer Perceptron Neural Network was used for recognizing intricate Braille patterns. A Long Short-Term Memory (LSTM) network facilitated the conversion of sequential Braille data into readable text or words by retaining long-term dependencies.

Reading Braille in Real-Time

The sensor was tested under varying pressure levels and touch conditions to assess its performance. It achieved a 98.57 % accuracy rate in recognizing eight distinct Braille patterns. The sensor also demonstrated its ability to function in dynamic environments, such as sliding over a Braille board. It accurately read entire Braille words and responded to pressure in under 0.1 seconds, even when minor variations in characters were present.

This system is far more precise than older Braille-reading technologies that might miss imperfectly pressed dots. The flexible optical fiber resonator detects very small pressure differences, and PDH frequency locking ensures stability and accuracy, even with changing light or power fluctuations. Machine learning further enhances the system, allowing it to recognize Braille despite minor errors or pressure variations.

Rui Min, Study Co-Author, Beijing Normal University

Min also worked with Lin Ma from Shenyang Aerospace University.

The researchers aim to further enhance the sensor for practical applications. Efforts include increasing durability, reducing production costs, and optimizing the sensor for integration into various devices. They are also enhancing the incorporation of machine learning models to handle more complex Braille reading assignments.

Journal Reference:

‌Wang, H., et al. (2024) Optical tactile sensor based on flexible optical fiber ring resonator for intelligent Braille recognition. Optics Express. doi.org/10.1364/oe.546873.