MIT researchers led by mechanical engineering professor Xuanhe Zhao have created a wristband that sticks to the skin and emits miniature ultrasound images of the wrist’s muscles, tendons and ligaments. An AI model, trained on human‑labeled scans, converts each frame into the exact positions of all five fingers and the palm. It stays glued for consistent imaging.

In live tests the band let a wearer wirelessly drive a robotic hand, mirroring gestures to play a simple piano melody or shoot a mini basketball into a desktop hoop. Users also manipulated virtual objects, pinching to zoom or rotate items on a screen. The demonstrations proved the wrist‑mounted sensor can replace bulky motion‑capture rigs.

The team aims to shrink the device to roughly a smartwatch size and expand training data to cover diverse hand shapes and gestures. A larger dataset could teach humanoid robots delicate tasks such as surgery or enable more natural VR interaction. Today the ultrasound wristband already delivers real‑time, high‑dexterity control without cameras. Developers can integrate the band via standard Bluetooth protocols.