Skywalker

Ultrasonic sensing and deep learning that predict each finger’s movement from the forearm, for prosthetic hands.

Skywalker started in 2018 in the Robotic Musicianship Lab at Georgia Tech, building on earlier work in the lab.

Ultrasonic sensing on the forearm and machine learning predict how each finger moves - the finger-by-finger control a prosthetic hand needs.

The demonstration in the video uses support vector machines to classify finger movements. I then worked on a convolutional neural network that tracks each finger continuously, aiming for the lowest latency and the closest match to the real movement.

The project was supported by a seed grant from the Georgia Tech Petit Institute.