New self-learning prosthetic arm controlled by owner's brain signals

"Initially, software will be universal, but we will adapt it to each specific artificial arm," said researcher Nikita Turushev.

Brooks Hays

Researchers train the software that powers their prosthetic arm prototype to pick up on the myoelectric signals sent from the brain to muscle to perform motions. Photo courtesy of Tomsk Polytechnic University

 

The most maneuverable prosthetic arms require the attachment of a traction belt to the owner's shoulder. In addition to being cumbersome, users must contort their body into unnatural positions to trigger certain motions. Researchers at Tomsk Polytechnic University in Russia are working on a prosthetic arm that learns from the user's brain signals and anticipates expected movements.
Scientists are perfecting a prototype. They say the final product will be able to perform the full range of motions of a healthy arm.
The human brain sends myoelectric signals to muscles to trigger an expected motion. Researchers have designed an algorithm to analyze myoelectric signals and anticipate the expected motion of the user.
"Initially, software will be universal, but we will adapt it to each specific artificial arm," researcher Nikita Turushev said in a news release.

The software's machine learning algorithm will enable the arm to copy and recognize the myoelectric signals and patterns specific to its owner.

Researchers are teaching the algorithm the myoelectric signals used by more than 150 study participants to control their healthy limbs. The scientists say they'll be ready to present their prototype and software within two years.