💬 AI Choker Restores Speech After Stroke

The device was developed at the University of Cambridge. The first sensor detects neck vibrations, while the second monitors emotional state via pulse impulses.

Researchers discovered that patients became highly exhausted even when attempting to speak. This led to the concept of using two "AI assistants" based on the GPT-4o mini.

The first agent accurately converts vibrations into simple, pre-trained phrases. The second, taking into account context, emotional state, and external conditions, expands them into coherent speech. For example, "We go to hospital" → "Let's go to the hospital for a doctor's appointment."

AI reduced the word error rate to 4.2% and the sentence error rate to 2.9%. After introducing the second interpretive agent, patient satisfaction increased by 55%.

According to Luigi Occhipinti, who led the project, many patients gradually recover their speech. That is why they do not require brain implants, but rather simple, user-friendly rehabilitation options.

@hiaimediaen