USC researchers announced Wednesday they have developed a robotic hand capable of teaching itself to play simple piano melodies by ear after just minutes of practice, a breakthrough scientists say could eventually help improve rehabilitation technology and treatment for movement disorders such as Parkinson’s disease.
The system, dubbed the “Musician Hand,” was developed by researchers at the USC Viterbi School of Engineering. The findings were published in the Journal of the Royal Society Interface.
Unlike traditional robots that rely heavily on pre-programmed instructions and massive training datasets, the robotic hand learned through a process researchers compared to the way infants experiment with movement while developing motor skills.
For roughly two minutes, the robot randomly pressed piano keys while recording the sounds produced and the finger movements needed to create them, according to USC researchers.
After the brief practice period, the hand was able to hear and reproduce an unfamiliar melody of about 30 notes without corrections or sheet music.
Researchers said the robotic system became advanced enough to perform before two music judges who listened blindly to recordings from the robot alongside performances by four human pianists.
“The Achilles’ heel of traditional robotics is the assumption that perfect information is necessary to act well,” Francisco Valero-Cuevas, a USC professor of biomedical engineering, aerospace and mechanical engineering, said in a statement. “Animals don’t work that way. They perceive; they guess, usually correctly; and they adapt. We wanted to show a robot could do the same.”
The hand uses four tendon-driven fingers controlled by small electric motors designed to mimic the mechanics of the human hand, while neural networks analyze melodies and convert them into finger movements.
Researchers said the project could eventually lead to advances in prosthetics, wearable robotic exoskeletons and physical therapy systems capable of adapting to how individuals naturally move.
“Imagine if, when you were first diagnosed, you wore an exoskeleton — a wearable robotic suit — and it learned how you move with only a few days of training,” Valero-Cuevas said in describing possible applications for Parkinson’s patients.
The research was supported by the National Science Foundation and the Defense Advanced Research Projects Agency.