Data Science Investment Counter —Funding raised by UK data science companies in 2018.
£ 5.640 Million

Artificial intelligence improves control of prosthetic hands

Scientists have created a new artificial intelligence-powered method for improving control of prosthetics that would drastically improve the technology’s efficiency.

The findings from The University of Texas at Dallas were shows at the 2019 IEEE International Symposium on Measurement and Control in Robotics last month.

To better understand in which way the new research is fundamentally innovative, it is necessary to look at the current state of prosthetics technology.

Natural muscles’ mobility function works through response to nerve impulses. Electromyography (EMG) is a diagnostic procedure that records the electrical impulses produced by muscles.

EMG is currently the most effective way of controlling prosthetic hands, and EMG signals have been used significantly in the recognition of users trying to control assistive technology devices like wheelchairs, prosthetic devices and exoskeletons.

Traditional results in prosthetics have improved greatly in the past few years using EMG, but they are still far from producing fully proficient, real-time movement.

The new study, called “Deep learning approach to control of prosthetic hands with electromyography signals“, describes a new method using artificial intelligence and deep learning to control prosthetic hands with raw EMG signals, meaning without preprocessing results. This would allow for a faster electric transfer between the brain and the muscles, and consequently faster hand movements for the user.

Researchers used real user data to re-train the system multiple times, modifying actions based on user requests. The new control system is based on the Python programming language and uses the TensorFlow deep learning library to run in real-time on an embedded general-purpose graphics processing unit developer kit.

“Our solution uses a novel deep convolutional neural network to eschew the feature-engineering step,” said Mohsen Jafarzadeh, lead researcher at The University of Texas at Dallas.

“Removing the feature extraction and feature description is an important step toward the paradigm of end-to-end optimization. Our results are a solid starting point to begin designing more sophisticated prosthetic hands.”

Image via wanicon on flaticon.

SHARE THIS ARTICLE:
BY SHACK15

Co-working space and blog dedicated to all things data science.

Subscribe to our newsletter