People who are deaf or cannot speak, or both – have long been aided by the American Sign Language (ASL) in communicating with people. Although the accuracy of the communication solely depends on the understanding of people on both ends, that can be a problem if one or more conversants are not that fluent in ASL. Researchers at Cornell University have recognized this problem and want to leverage the goodness of wearable technologies to make it smoother for people who are differently abled to communicate with the world as they intend to.
The experts at the university have employed the smart ring platform to develop a wearable that tracks the American Sign Language gestures of the wearer to accurately transcribe them into understandable text or audio to a computer or a smartphone. They call it the SpellRing, and it can be worn on the thumb to get the desired input for the translated data. The wearable comes with embedded sensors including a mini-gyroscope to measure the angular data from hand movement. The on-board speakers and mic detect the inaudible sound waves from the movement. All these components are housed inside a 3D-printed housing that’s no more than the size of a quarter.
Designer: Cornell University
The smart AI ring can detect all 26 letters in the English language with micro-sonar technology to precisely track the fingerspelling of the wearer. For more accuracy sonar images are also fed into the system. The data that is received from all the sensors and hardware is then processed by a deep-learning algorithm. All of this is done in a split second time to predict the ASL fingerspelled by the wearer. To make possible an in-depth and unbiased view of the wearable for further development, the researchers from Carnegie Mellon University and Stony Brook University, along with the researchers from Cornell University have tested the wearable thoroughly.
In the conducted tests, 20 amateur and experienced ASL signers were asked to spell more than 20,000 words to see the accuracy of the AI-assisted wearable ring. The 3D-printed wearable passed with a combined accuracy of around 90 percent with both signed words, as well as, sentences. It was even tested to perform actions such as web search, map navigation, or typing in notes with the input. According to Hyunchul Lim, lead author of SpellRing’s papered documentation, “ASL is a very complicated, complex visual language,” and the focus is right now on tuning the algorithm to make the AI ring understand words and phrases with maximum accuracy.
The SpellRing is currently in the prototype stage and there is no word on the availability of the ring yet. The team is confident that even in the current stage, ASL fingertyping will be significantly faster than typing on your smartphone’s keyboard. Researchers want this ring to be a tool for fast and accessible text entry with a niche set of benefits targeted toward a set of users. The prospects of this AI ring graduating to the wearable market are very high since it costs just $30 to make the current prototype AI ring. When it will be mass produced that cost can be further trimmed down.
The post SpellRing is an AI-enabled smart ring that converts American Sign Language into phone text first appeared on Yanko Design.