What is it about?
Speaking is not only about sound. Even when a person mouths words silently, the skin and muscles around the neck move in tiny but meaningful patterns. In this study, we developed a soft wearable neckband that can read these subtle movements. The device uses a small camera and visual markers on a flexible silicone sensor to create a movement map of the throat while a person silently forms words. An AI model then analyzes these patterns to identify the intended word and can connect the result to a voice synthesis system trained on the user’s own voice. Because the system reads neck movement rather than sound, it can work in loud places where ordinary microphones struggle. In early tests, it recognized the 26 NATO phonetic alphabet words, such as “Alpha” and “Bravo,” and continued to perform in very noisy conditions. This work points toward future communication tools for people with voice disorders and for workers, responders, or others who need to communicate clearly when speaking out loud is difficult or impossible.
Featured Image
Photo by Balint Mendlik on Unsplash
Why is it important?
This work is important because it offers a new way to communicate when sound-based speech is unavailable, unreliable, or unsafe. Many existing silent speech systems rely on bulky equipment, single-use electrodes, or unstable sensors that capture only limited movement information. Our approach uses a soft optical strain sensor worn on the neck to capture two-dimensional movement patterns of the skin and muscles involved in speech. This richer “movement map” provides the AI with more useful information for decoding silent speech, while eliminating the reliance on a microphone. The system was tested in extremely noisy conditions including noise levels comparable to industrial or construction environments and was designed for real-time use on compact computing hardware. With further development, this technology could help people who have lost their natural voice due to illness or surgery. It could also support clearer communication in factories, emergency response, aviation, maritime work, military operations, and other environments where normal speech or microphones may fail.
Perspectives
This research focused on bringing silent speech technology closer to everyday use by making it soft, wearable, noise-robust, and capable of preserving a person’s own vocal identity rather than producing a generic machine voice. While the current system is an early step focused on a limited but practical alphabet-based vocabulary, our future goals are to expand it toward sentence-level communication, improve comfort and robustness during long-term use, and test the system with more diverse users. I hope this research will contribute to future communication systems that help people communicate more naturally in hospitals, workplaces, emergency settings, and everyday life.
Sunguk Hong
Pohang University of Science and Technology
Read the Original
This page is a summary of: Soft Multiaxial Strain Mapping Interface with AI-Driven Decoding for Silent Speech in Noise, Cyborg and Bionic Systems, January 2026, Tsinghua University Press,
DOI: 10.34133/cbsystems.0536.
You can read the full text:
Contributors
The following have contributed to this page
