Show HN: Subvocal Introduces Smart Glasses for Silent Computer Interaction

A pair of computer science students turned founders have introduced an innovative wearable device designed to let users communicate with computers without uttering a single audible word. Developed by Aarush and Arnav from The Subvocal Company, the smart glasses utilize advanced radio frequency technology to decode silent mouthing and internal speech into written text at normal conversational speeds.

The project originated from a desire to create a practical solution for using voice-activated tools and AI assistants in public spaces like libraries or shared offices, where speaking out loud is impractical. Initially experimenting with facial electromyography sensors during a hackathon, the creators quickly realized that traditional muscle-sensing methods faced significant hurdles due to sweat, electrical interference, and the impracticality of attaching electrodes to the face.

To overcome these challenges, the team transitioned to radio frequency waves that pass through the user’s face. As micro-movements of the tongue, lips, and soft palate modulate the waves, foundational machine learning models process the resulting perturbations and translate them accurately into text. Developing the hardware required extensive simulation and custom antenna integration directly within the frames of stylish eyewear.

Addressing cross-user anatomical differences required massive data collection efforts, leading the founders to train their models on diverse datasets gathered from volunteer participants. With the technology now reaching a refined stage, the company has opened preorders for an initial batch of units, marking a significant step toward practical silent speech interfaces.

Sources: Hacker News

Aggregated summary. Original: Hacker News