AI Neckband Lets People Speak Without Making a Sound

Imagine being able to talk to someone across a noisy factory floor without making a sound, or helping someone who lost their voice to cancer speak again in their own tone and inflection.

Scientists at Pohang University of Science and Technology in South Korea just made that possible with a simple silicone neckband that reads the invisible language of your throat.

Every time you form a word, your neck muscles and skin shift in subtle but predictable ways. Think of it as a silent fingerprint for each syllable. The POSTECH team figured out how to read those tiny movements and turn them into actual speech.

The neckband works by tracking how your skin stretches and in which direction when you mouth words silently. A miniature camera reads reference markers printed on the silicone collar, measuring those deformations in real time. An AI model then identifies which word you articulated and sends it wirelessly to a server that synthesizes it as audio in your own voice.

Training the system takes less than 10 minutes of recordings. After that, it reproduces your personal intonation and vocal character with remarkable accuracy.

In tests using the NATO phonetic alphabet (words like "Alpha," "Bravo," and "Charlie"), the system achieved 85.8% accuracy. The collar even held up against background noise as loud as a construction site, outperforming commercial systems that measure electrical muscle activity.

Why This Inspires

Professor Sung-Min Park, who led the research, points to patients who've lost their voice to surgery or disease as the primary beneficiaries. For someone who had their larynx removed, this could mean communicating again in something close to their original voice rather than through a mechanical device.

The applications stretch far beyond medicine. Factory workers could communicate clearly over machinery noise. Emergency responders could coordinate silently during critical operations. Pilots and maritime crews facing extreme noise conditions could maintain clear communication.

The system does have current limitations. It works only with 26 predefined words, not free conversation, and accuracy drops when users walk or make big head movements. The team is now working to expand the vocabulary, test with more users, and improve performance during physical activity.

This follows earlier work from Cambridge researchers who achieved 95% accuracy with a similar throat sensor and even added emotional state detection. The POSTECH team's innovation of recreating the user's actual voice adds a deeply personal dimension to the technology.

For the 7.5 million Americans with voice disorders and countless others working in impossibly loud environments, silent speech technology is moving from science fiction to practical reality.