MIT Wristband Uses Ultrasound and AI to Capture Hand Motion for Robot Training
Researchers developed an ultrasound wristband that records muscle and tendon movements beneath the skin. The device transmits images to an AI system that translates the data into commands for a robotic hand.
Abc NewsResearchers at the Massachusetts Institute of Technology have created an ultrasound wristband that records muscle, tendon, and ligament motion to generate training data for robotic hands. The wristband sends high-frequency sound waves through the wearer’s skin and relays the resulting images to a computer.
An AI algorithm converts those images into the 22 degrees of freedom that describe human hand movement.
In tests with eight volunteers, the system reproduced all 26 letters of American Sign Language and other gestures within 120 milliseconds. The wristband operates wirelessly, allowing the operator and robot to be in separate locations. Developers said the collected data could be used to train robots for tasks such as housework or surgical procedures that require fine finger control.
The team also envisions building large motion datasets that would let humanoid robots learn dexterous actions without ongoing human input.
