AI Rings Are Breaking Down Communication Barriers

Communication is a challenge for those who have a speech or hearing impairment. These individuals can express themselves through sign language; however, relatively few people understand it. Many attempts have been made to bridge this knowledge gap with assistive technologies, such as computer vision systems or gloves that can translate signs into spoken language. While these systems work well in certain settings, they tend to be bulky or uncomfortable, making them unsuitable for regular use.

A more practical solution to this problem has just been introduced by a team led by researchers at Yonsei University. They have developed a ring-based device that can recognize the finger gestures of sign language and translate them into speech in real-time. Unlike previous approaches, these rings are small, lightweight, and practical for all-day use.

The new system, called the wirelessly connected ring-type sign language translator (WRSLT), replaces the traditional glove design with several independent sensor rings worn on selected fingers. Each ring contains an accelerometer that tracks finger movement and hand orientation. The data is then transmitted wirelessly to a processing system that interprets the gestures using artificial intelligence.

Conventional wearable sign language translators often rely on gloves fitted with multiple wired sensors. While functional, these systems can restrict natural movement and become uncomfortable during prolonged use. Fixed sensor placement inside gloves can also create problems for users with different hand sizes and finger shapes, reducing accuracy and requiring individual calibration.

The researchers addressed these issues by designing fully wireless rings that operate independently without cables connecting them together. Because the rings can be positioned freely on the fingers, the system adapts more naturally to different users. The modular design also improves comfort and allows for more fluid hand movement during dynamic signing.

To simplify the hardware without sacrificing performance, the team conducted an analysis to determine which fingers contributed most to accurate sign recognition. From this study, they identified seven key fingers that provided the strongest classification signals, reducing the number of required sensors while maintaining high accuracy.

The WRSLT system was trained to recognize both American Sign Language and International Sign Language. In testing, it achieved recognition accuracies of 88.3 percent for American Sign Language and 88.5 percent for International Sign Language when evaluated on users who were not included in the training dataset. This unseen user performance is particularly important because it demonstrates that the system can generalize to new users without requiring personalized retraining.

The device can do more than just recognize individual words. The researchers developed a sequential word detection framework capable of translating continuous signing into sentence-level output in real-time. Instead of requiring pauses between words or manually segmented gestures, the system can process natural streams of signing as connected language.

The researchers believe their technology could eventually be integrated into a commercial device to help facilitate communication between signers and nonsigners in daily life. That would be a big win for those who rely on sign language.